CN117934035A

CN117934035A - Method, device and storage medium for predicting construction cost of building construction

Info

Publication number: CN117934035A
Application number: CN202410088952.7A
Authority: CN
Inventors: 黄徐斌; 张涵; 何旦聃; 庞菲菲
Original assignee: Suzhou Huaxing Engineering Cost Consulting Co ltd
Current assignee: Suzhou Huaxing Engineering Cost Consulting Co ltd
Priority date: 2024-01-22
Filing date: 2024-01-22
Publication date: 2024-04-26
Anticipated expiration: 2044-01-22
Also published as: CN117934035B

Abstract

The application discloses a method and a device for predicting construction cost of a building construction project and a storage medium, and belongs to the technical field of computers. The method is carried on a construction cost estimation platform at a server end or a terminal, when the method is used for carrying out prediction, internal and external feature engineering data of an engineering project to be predicted are obtained, a database analysis model is input to obtain similar engineering groups which are accurately matched with the engineering data, index collection calculation is carried out to obtain meta engineering quantity index data of each engineering project, combination calculation is carried out on each meta engineering data contained in each engineering project and the matching degree of corresponding engineering levels of the meta engineering data to obtain initial data of each meta engineering quantity index contained in the engineering project to be predicted, and a prediction difference adjustment model is input by combining the meta engineering quantity index data of each engineering project together to obtain engineering quantity index correction coefficients corresponding to each meta engineering data contained in the engineering project to be predicted; the problems of low project cost data standardization degree, huge data screening processing capacity and difficult index establishment in the related technology can be solved by accurately matching similar project engineering groups and analyzing and calculating meta engineering data, and compared with the method relying on artificial experience, the project cost index can be rapidly and accurately generated and predicted by a computer, so that the management benefit and project investment management and control are improved, and the resource investment is saved and the social and economic benefits reach the standard.

Description

Method, device and storage medium for predicting construction cost of building construction

Technical Field

The embodiment of the application relates to the technical field of computers, in particular to a method and a device for predicting construction cost of a building construction and a storage medium.

Background

The house construction project outline refers to the process of reasonably planning and estimating various resources required by the project before the building project starts, and is also an important means for project cost control.

The construction cost consulting result of the traditional house construction project depends on experience summary obtained by consultants according to past historical data or an average condition of the historical data, but the requirements of quick valuation, rechecking and calculation under the situation that design drawing data of a construction cost consulting enterprise in a project early consultation stage is insufficient or the depth requirement is not met exist, and the construction cost estimating method of the house construction project in the traditional mode is difficult to realize the requirements.

Therefore, in order to meet the increasing and complex estimated demands of the cost index, the traditional processing mode needs to be changed, the processing efficiency is further improved, and valuable cost index data are ensured to be obtained.

Disclosure of Invention

The embodiment of the application provides a method, a device and a storage medium for predicting construction cost of a building construction project, which are used for solving the prediction requirement of construction cost indexes and comprise the following steps:

in one aspect, a method for estimating construction cost of a building is provided, the method comprising:

acquiring internal and external characteristic engineering data of project engineering to be predicted;

inputting the internal and external feature engineering data of the project to be predicted into a database analysis model, and determining similar project engineering groups matched with the internal and external feature engineering data of the project to be predicted, wherein the similar project engineering groups are formed by at least one matched project engineering and are sequentially arranged according to project engineering level matching degree, and the project engineering indicates the included meta engineering quantity index data through meta engineering data;

performing index collection calculation to obtain meta-engineering quantity index data of each engineering;

Performing combined calculation on each piece of meta-engineering data contained in each piece of project engineering and the corresponding project engineering level matching degree to obtain initial data of each piece of project engineering quantity index contained in the project engineering to be predicted;

Inputting each piece of meta-engineering data contained in each piece of project engineering and each piece of meta-engineering quantity index initial data contained in the project engineering to be predicted into a prediction difference adjustment model to obtain engineering quantity index correction coefficients corresponding to each piece of meta-engineering data contained in the project engineering to be predicted;

and generating a house construction project cost estimation report corresponding to the project to be predicted according to the project quantity index correction coefficient corresponding to each piece of meta project data contained in the project to be predicted and each piece of meta project quantity index initial data contained in the project to be predicted.

Optionally, before the obtaining the internal and external feature engineering data of the project engineering to be predicted, the method further includes:

acquiring historical engineering project data, and constructing the database analysis model according to the historical engineering project data;

And constructing the meta-engineering data from the historical engineering project data according to a metering pricing rule.

Optionally, the building the meta-engineering data from the historical engineering project data according to the metering rule includes:

According to the metering and pricing rule, content division is carried out on each historical engineering project from the historical engineering project data to obtain each meta engineering;

and carrying out automatic collection on engineering quantity list items belonging to the same meta-engineering to obtain meta-engineering data corresponding to each meta-engineering.

Optionally, after the data of the internal and external feature engineering of the project engineering to be predicted is input into a database analysis model, the method further includes:

acquiring main internal and external feature item names and corresponding values in internal and external feature engineering data of the project engineering to be predicted, wherein the main internal and external feature item names are obtained according to preset selection;

Matching corresponding similar project engineering groups from the historical engineering project data according to the names of the main internal and external feature items and the corresponding numerical values, and performing aggregation calculation;

and feeding back meta-engineering data of each project in the similar project engineering group after calculation is completed.

adding a custom collection label to each meta-project in the similar project group;

judging and prompting the correctness of the added custom aggregation label according to the consistency requirement of the meta engineering and the corresponding metering unit;

And updating and exporting the similar project engineering groups.

Optionally, the method further comprises:

Acquiring input project key feature information about project engineering data to be predicted;

Inputting the internal and external feature engineering data of the project to be predicted into a database analysis model, and determining similar project engineering groups matched with the project engineering data to be predicted, wherein the method comprises the following steps:

Carrying out standardized processing on the project characteristic key information, and inputting the project characteristic key information into the database analysis model;

and matching similar project engineering groups corresponding to the project characteristic key information through the database analysis model.

Optionally, the inputting the metadata included in the project projects and the initial data of the metadata amount indicators included in the project projects to be predicted into a prediction adjustment model to obtain engineering amount indicator correction coefficients corresponding to the metadata included in the project projects to be predicted, includes:

Inputting meta-engineering quantity index data contained in each engineering in the similar engineering group and initial data of each meta-engineering quantity index contained in the engineering to be predicted into a prediction difference adjustment model to obtain the matching degree of meta-engineering layer level of each meta-engineering quantity index data contained in each engineering in the similar engineering group;

and calculating and averaging the matching degree of the meta-engineering layer level of each meta-engineering quantity index data contained in each engineering in the similar engineering group to obtain the engineering quantity index correction coefficient of each meta-engineering data contained in the engineering to be predicted.

Optionally, the engineering quantity index correction coefficient is recorded as Δw _x, and the calculation formula is as follows:

Wherein w _ji (N) is the synaptic weight of the neuron i connected to the neuron j, deltaw _ji (N) is the corrected value of the synaptic weight w _ji (N) trained for N times, and N is a positive integer, and N is more than or equal to 1 and less than or equal to N.

On the other hand, still provide a house construction engineering cost and predict device, the device includes:

The data acquisition module is used for acquiring the internal and external characteristic engineering data of the project engineering to be predicted;

the data input module is used for inputting the internal and external feature engineering data of the project engineering to be predicted into a database analysis model, and determining similar project engineering groups matched with the internal and external feature engineering data of the project engineering to be predicted, wherein the similar project engineering groups are formed by at least one matched project engineering and all project engineering are sequentially arranged according to project engineering level matching degree, and the project engineering indicates the contained meta engineering quantity index data through meta engineering data;

the collection calculation module is used for executing index collection calculation to obtain meta-engineering quantity index data of each engineering;

the combined calculation module is used for carrying out combined calculation on each piece of meta-engineering data contained in each piece of engineering and the corresponding item engineering level matching degree to obtain each piece of engineering quantity index initial data contained in the project to be predicted;

the prediction difference adjustment module is used for inputting each piece of meta-engineering data contained in each piece of project engineering and each piece of meta-engineering quantity index initial data contained in the project engineering to be predicted into a prediction difference adjustment model to obtain engineering quantity index correction coefficients corresponding to each piece of meta-engineering data contained in the project engineering to be predicted;

The construction cost estimation module is used for generating a house construction project construction cost estimation report corresponding to the project to be predicted according to the project quantity index correction coefficient corresponding to each piece of project data contained in the project to be predicted and each piece of project quantity index initial data contained in the project to be predicted.

In another aspect, a server is provided, the server comprising a processor and a memory; the memory stores at least one instruction for execution by the processor to implement the method of predicting construction costs of a building as described in the above aspect.

In another aspect, a computer readable storage medium is provided, the storage medium storing at least one instruction for execution by a processor to implement the method of building construction cost estimation as described in the above aspect.

In another aspect, there is provided a computer program product storing at least one instruction that is loaded and executed by the processor to implement the method of predicting construction cost of a building according to the above aspect.

The prediction method of the construction cost of the building engineering provided by the application can at least provide the following technical effects.

The method is carried on a construction cost estimation platform at a server end or a terminal, when the method is used for carrying out prediction, internal and external feature engineering data of an engineering project to be predicted are obtained, a database analysis model is input to obtain similar engineering groups which are accurately matched with the engineering data, index collection calculation is carried out to obtain meta engineering quantity index data of each engineering project, combination calculation is carried out on each meta engineering data contained in each engineering project and the matching degree of corresponding engineering levels of the meta engineering data to obtain initial data of each meta engineering quantity index contained in the engineering project to be predicted, and a prediction adjustment model is input together with the meta engineering quantity index data of each engineering project to obtain engineering quantity index correction coefficients corresponding to each meta engineering data contained in the engineering project to be predicted; the problems of low project cost data standardization degree, huge data screening processing capacity and difficult index establishment in the related technology can be solved by accurately matching similar project engineering groups and analyzing and calculating meta engineering data, and compared with the method relying on artificial experience, the project cost index can be rapidly and accurately generated and predicted by a computer, so that the management benefit and project investment management and control are improved, and the resource investment is saved and the social and economic benefits reach the standard.

Drawings

FIG. 1 is a flow chart illustrating a method for predicting construction costs of a building according to an exemplary embodiment of the present application;

FIG. 2 is a schematic diagram of a construction project cost estimation system corresponding to FIG. 1;

FIG. 3 illustrates an interface diagram characterizing the degree of matching between similar project projects and project projects to be predicted within a database;

FIG. 4 is a flow chart illustrating a method for predicting construction costs of a building according to another exemplary embodiment of the present application;

FIG. 5 shows a schematic diagram of an induced local area v _j (n) generated at the activation function input of neuron j provided by an exemplary embodiment of the present application;

FIG. 6 is a schematic diagram illustrating a process of all instantaneous errors E _n of a network according to an example embodiment of the present application;

FIG. 7 illustrates a schematic structure of three samples of a predictive robust model;

fig. 8 is a block diagram illustrating a construction of a construction project cost estimating apparatus according to an exemplary embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the embodiments of the present application will be described in further detail with reference to the accompanying drawings.

References herein to "a plurality" means two or more. "and/or", describes an association relationship of an association object, and indicates that there may be three relationships, for example, a and/or B, and may indicate: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship.

Based on the existing cost estimation problem proposed by the background technology, the accumulation and application of historical project data obtained by analysis of the applicant are one of key factors for solving the problem, and the historical project data and corresponding project information accumulated by the enterprise are required to be effectively utilized to form valuable cost index data.

Meanwhile, the working scheme is reformed for actively responding to the construction cost of the building. The construction cost database for accelerating the establishment of national capital investment is published manually, materials, projects and other construction cost index indexes according to the types of areas, engineering types and building structures, and the like, and large data, artificial intelligence and other informatization technologies are utilized to provide basis for budget estimate planning, so that the construction cost consultation enterprises with the capability are encouraged to establish and perfect the enterprise construction cost database.

In addition, the design thought disclosed by the construction project cost prediction method can be widely applied to scenes needing to be subjected to cost prediction, such as other production fields. For convenience of explanation of the following embodiments, each embodiment of the present application is schematically illustrated by taking estimated building construction cost index as an example, but the present application is not limited thereto.

Referring to fig. 1, a flowchart of a method for estimating construction cost of a building according to an exemplary embodiment of the application is shown. In addition, fig. 2 shows a schematic diagram of a construction project cost estimation system corresponding to fig. 1. The method shown in fig. 1 comprises the following steps:

and step 101, acquiring internal and external characteristic engineering data of project engineering to be predicted.

The internal and external feature engineering data of the project to be predicted are the internal and external feature engineering data of the project to be predicted corresponding to the house building project to be evaluated at present, in order to facilitate the input of a database analysis model, the internal and external feature engineering data of the project to be predicted can be processed data required by the input of a matching model, and can also be data files related to the internal and external feature engineering of the project to be predicted, and the internal and external feature engineering data of the project to be predicted, which are matched with the input database analysis model, are obtained through the arrangement of technicians or the arrangement of electronic documents.

In one possible implementation, the input of the model includes characteristics of the project, such as construction area, fortification intensity, structure type, foundation form, building height, floor area, building area, number of floors of the ground, and the like. The output is the estimated cost of the project. The model is trained through historical data, and accuracy of the model on a new project is ensured. The database analysis model is a key component in the whole house construction project cost estimation method. The following is an exemplary description of database analysis model construction, training, and technical features.

In the process of constructing a database analysis model, model selection is firstly performed. A database analysis model is selected that suits the problem. Machine learning models, such as regression models, may be considered to predict the cost of a project. Another option is to use a deep learning model, such as a neural network, to handle more complex data patterns. Further, feature selection is performed. Features to be used to construct the model are determined. Features may include internal and external features of the project, such as construction area, fortification intensity, structure type, foundation form, building height, floor area, building area, number of floors of land construction, etc. Ensuring that the selected feature has an impact on the cost of the project and is able to provide sufficient information for accurate estimation. In addition, data preparation is performed. A database for training is collected and prepared. This may include details of the history item including the number of individual projects, unit price, etc. Ensure the quality of data and carry out necessary cleaning and pretreatment.

In the database analysis model training process, data is firstly divided into a training set and a testing set so as to ensure that the model fully learns the data mode. And then, training the training set by using a database analysis model so that the training set can understand the mode and the association of the historical engineering project data. Finally, the model is simple and optimal, and the generalization performance of the model on new projects is improved mainly by adjusting super parameters. It should be noted that the current module only realizes the two functions of automatic data collection to meta engineering data and project engineering level matching degree calculation, so that the realization of model training and tuning is simplified. In practical cases, researchers can further optimize as needed to fully exploit the potential of database analysis models.

In one example, model training and tuning may be further refined, such as dividing the data set into a training set and a test set. The training set is used for training of the model, and the test set is used for evaluating the performance of the model. Generally, the division may be performed with a ratio of 80-20 or 70-30. And secondly, training a model, and training a training set by using the selected database analysis model. During the training process, the model will learn patterns and associations in the data so that new projects can be accurately predicted later. Finally, performing model tuning: and adjusting the super parameters of the model according to the performance of the training set. This may be achieved by cross-validation or the like techniques to ensure the generalization performance of the model.

In the embodiment of the application, the database analysis model has the characteristics of machine learning model characteristics, deep learning model characteristics, a certain interpretation and transparency, generalization performance and the like.

The machine learning model adopts linear regression, decision tree, random forest and other algorithms. These models are generally applicable to problems with clear feature-target relationships, but may require more feature engineering data to improve accuracy. The foregoing is an exemplary illustration of the embodiments of the present application with respect to database analysis model construction, training, and characterization, the specific selection and implementation being adaptable to specific needs of the skilled artisan.

And 102, inputting the internal and external feature engineering data of the project to be predicted into a database analysis model, and determining similar project engineering groups matched with the internal and external feature engineering data of the project to be predicted.

The similar project engineering group is composed of at least one matched project engineering, and the project engineering is sequentially arranged according to the project engineering level matching degree. Wherein each project indicates the included meta-engineering quantity index data through meta-engineering data.

Firstly, based on related theoretical basis such as engineering cost index concept, and the like, combining with metering rule to establish minimum basic unit applied by engineering cost index, and marking as meta engineering, and marking each level of data contained under the meta engineering as meta engineering data. In one possible implementation, when the metadata is extracted from the "building construction and decoration engineering amount calculation Specification", how to map the engineering amount calculation rule in the Specification to a specific metadata item is explained by the following, so as to ensure that the data extracted from the Specification matches with the actual item.

And secondly, performing standard interpretation. The fine study is conducted on the "house construction and decoration engineering quantity calculation Specification", and the definition, calculation rules, application rules of measurement units and the like of different hierarchical engineering quantities are understood. Particular attention is paid to the relevant application specification of the individual engineering quantity items involved in the specification.

And secondly, performing meta engineering definition. The definition of the meta-engineering is determined in the specification. The meta-engineering may be a class of engineering defined explicitly in the specification, may be a basic building unit, or may be some kind of common engineering component, ensuring a clear understanding of the definition of the meta-engineering.

And further performing engineering calculation rule mapping. The engineering quantity calculation rules in the specification are mapped to specific meta-engineering data items. This may require the creation of a rule mapping table that corresponds the calculation rules for each engineering quantity definition item in the specification to the meta-engineering data item. The accuracy and the comprehensiveness of rule mapping are ensured. In the scheme, a mapping rule is established in combination with existing rules of the specification, and the first 9 digits in the 12-digit codes corresponding to the minimum metering unit engineering quantity list items defined in the specification are mapped to the same metadata engineering data item.

The specific steps include identifying meta engineering, formulating a mapping table, considering detail calculation, unit standardization and verifying mapping relation. Identifying a well-defined meta-project from the specification, involving chapters or appendices in the specification, classifying and describing the meta-project; establishing a mapping table, listing each engineering quantity calculation rule in the specification and specific meta-engineering data items corresponding to each engineering quantity calculation rule, wherein the mapping table comprises detailed information such as a calculation formula, a unit, a calculation rule and the like in the specification; in the mapping process, the details of engineering quantity calculation are considered, for example, if a calculation rule of a certain building material is involved in the specification, the fact that the mapped meta-engineering data item can accurately express the usage quantity of the material is ensured; ensuring that the mapped meta-engineering data item accords with corresponding unit standards, and ensuring that the mapped meta-engineering data item accords with the standards by the standard measurement units of various engineering quantities in the specifications; verification is performed using the actual project data. And selecting some actual projects, calculating meta-engineering data by using a mapping relation, and comparing the meta-engineering data with engineering quantity data of the actual projects to ensure the accuracy and reliability of mapping.

The combination specification provides for establishing mapping rules, wherein one rule is to map the same first 9 digits in the 12-digit codes corresponding to the minimum metering unit engineering quantity list items defined in the specification to the same meta-engineering data item. The mapping procedure is described by the following example.

Assume that the actual pricing file has the following two minimum metering unit engineering quantity items listed: 1. the beam plate is arranged, the coding is 010505001004, the engineering quantity is 103.55m ³, and the amount is 79563.3464 yuan; 2. liang Xieban, code 010505001005 engineering amount 2.77m ³ money 1745.4921 yuan.

In both encodings, the first 9 digits 010505001 are identical, and they will be mapped into the same metadata item according to the mapping rules. During the mapping process, the system integrates the relevant information of the two minimum metering unit engineering quantity items into one meta engineering data item.

The mapped meta-engineering data item includes at least the following information. Meta-engineering name: a beam plate is arranged; coding range: 010505001 (the first 9 identical parts); related metering information: the engineering quantity is 106.32m ³ and the amount of money is 81308.8385 yuan.

Through such mapping rules, the system can effectively categorize the minimum metering unit engineering quantity list item with the matching degree into the same meta-engineering data item, thereby providing a more organized data structure for subsequent analysis and prediction. This helps to simplify data management and improve efficiency. In order to ensure the validity of the mapping relationship, the updating of the specification needs to be checked regularly, so as to ensure that the mapping relationship is still applicable to the latest version of the specification. In addition, the mapping relation can be adjusted according to the specific characteristics of the project. The flexibility of the mapping relation is ensured, and the method can be suitable for different types of building engineering projects.

In the embodiment of the application, the metering and pricing rules include the existing "house building and decoration engineering quantity calculation Specifications" GB 50854-2013, and "construction engineering quantity list pricing Specifications" GB 50500-2013, etc., which are not limited herein.

It should be noted that the meta engineering is not completely consistent with the hierarchical engineering concept in the existing engineering cost field. In the prior art, the grading engineering in the engineering cost field comprises engineering quantity clearing items, sub engineering, unit engineering and single engineering according to grades from small to large.

Further, by utilizing an intelligent algorithm, a plurality of similar project engineering groups which are accurately matched with project engineering data to be predicted are automatically matched and determined in a large number of project engineering which are recorded in a database and complete information maintenance. In one example, the intelligent algorithm described above selects a classification algorithm in machine learning, such as a support vector machine (Support Vector Machine, SVM), and applies it to the scene of item matching. First, a Support Vector Machine (SVM) is selected as a classification algorithm for item matching. The SVM is suitable for a high-dimensional data set, and can find an optimal hyperplane in a feature space to realize data classification. Second, key features are extracted from project engineering data, which may include engineering quantity, cost, engineering duration, etc. for each level. And (3) performing dimension reduction on the features by using Principal Component Analysis (PCA) to reduce redundant information. The features are further standardized to ensure that they are on the same scale. This helps to improve the training effect of the algorithm.

In addition, preparation of training data is required. Historical project engineering data is prepared as a training data set, and the matching degree of the historical project engineering data is marked as a training label. It is ensured that the dataset contains a combination of various project engineering features to enhance the generalization ability of the algorithm. Model training is performed on the training data using a support vector machine. The SVM will learn the relationship between the history project engineering features and the degree of matching. The trained model was evaluated using the validation dataset to check its performance on unseen data. The evaluation index may include accuracy, recall, precision, etc. And performing super-parameter tuning according to the evaluation result to further improve the model performance. And applying the trained support vector machine model to real-time item matching. And inputting characteristic data of project engineering to be predicted, and outputting the matching degree through a model. And (5) periodically collecting an actual matching result, and comparing the actual matching result with the model output. And updating and improving the model according to the comparison result so as to adapt to new data distribution. And finally, deploying the trained and evaluated support vector machine model into an actual system for automating the project matching process.

In another example, the intelligent algorithm described above selects a clustering algorithm in machine learning, such as a k-means algorithm, and applies it to a scene of item matching. The matching degree calculation is the core of cluster analysis, and the matching degree between individuals is measured by distance according to different matching degree calculation modes of clustered objects, and common distance calculation modes comprise Euclidean distance, absolute value distance and Minkofski distance.

In view of the diversity of construction projects, various different types of data need to be processed. In order to more accurately represent and process such data, different types of variables are introduced in embodiments of the present application. Firstly, in the training process of a database analysis model, historical engineering project data is collected and arranged. These data include rich information, some of which are interval variables such as building height, assembly rate, etc. These interval variables help model learn patterns and associations in the data in the training of the model. On the other hand, two-state variables are also involved, i.e. variables with only two possible values. In project data, there are certain features that are binary, such as whether the nature of construction is new or rebuilt. By introducing two-state variables, the impact of these binary features on cost can be captured more accurately. In addition, classification variables are also considered to describe the category or group of items. For example, green building level may be considered a classification variable, including one star, two star, three star, etc. The method is beneficial to operators to carry out more detailed analysis on building projects with different green building grades. Finally, the concept of ordinal variables is introduced, a variable that describes ordered categories. In a building construction, one example is the earthquake resistance level of a building, which is an ordered classification variable. With these relationships in mind, operators can better understand the impact of different shock levels on cost.

Thus, by taking these different types of variables into account, an operator can more fully understand the characteristics of the project data, providing more accurate and valuable information for model training and prediction. The selection and use of these variables can be better adapted to the characteristics and complexity of the different projects. Specific calculation contents of the interval variable, the two-state variable, the classification variable and the ordinal variable are described below.

1. Interval variable. Is a continuous variable such as length, width, height, etc. Two individual objects a and b are provided, n interval variable values describing the individual object a are x _a1,x_a2,x_a3…,x_an, n interval variable values describing the individual object b are x _b1,x_b2,x_b3…,x_bn, and the distance (difference degree) between the individual objects a and b is d _ab. The euclidean distance formula between the two is shown as formula (formula 1).

2. Two state variables. Is a discrete variable, and the values are generally 0 and 1. The difference degree of the two-state variable is calculated in a different mode from the interval variable, and corresponding statistics is usually needed to be carried out on the values of the two-state variable before the difference degree calculation is carried out. Taking the two individual objects a and b as examples, n binary variables are described as the objects a and b, and the statistical condition of the values of the binary variables in a and b is shown in the following table 1.

TABLE 1

The number c of cases where the binary variable value in the object a is 1 and the binary variable value in the object b is 0, the number d of cases where the binary variable value in the object a is 1 and the binary variable value in the object b is 1, the number e of cases where the binary variable value in the object a is 0 and the binary variable value in the object b is 1, the sum of the four cases f and c, d, e, f is n, and the difference calculation formulas of the objects a and b are as follows. a. The difference degree of the two variables b is the ratio of the number of the two-state variables in a and b with different values to the total number of the two-state variables.

3. Classifying the variables. The variable is also a discrete variable, but has 3 values and more than 3 values, and the classification variable is consistent with the idea of a two-state variable. Taking two individual objects a and b as examples, describing n classification variables of the objects a and b, and setting the number of cases when the classification variables in the object a are the same as the classification variable values in the object b as c, the difference degree calculation formulas of the objects a and b are shown as (formula 3). In most cases, the classification variable can be converted into an asymmetric two-state variable, and the two-state variable is calculated by using a difference degree calculation formula.

4. Ordinal variables. It may be a discrete variable or a continuous variable, and if the ordinal variable is a discrete variable, the ordinal variable is also generally classified but there is a certain order relationship between the classes. The calculation of the ordinal variable difference is obtained by normalizing the grade value of the variable, converting the grade value into an interval variable and calculating the interval variable difference. Taking the kth variable of the a object as an example. The normalized calculation formula of the rank value is shown below. Wherein R _ak is the rank value of the kth variable of the a object, R _k is the maximum value of the rank values divided by the kth variable of all objects, and Z _ak is the rank value after normalization, and the corresponding calculation formula is shown in formula 4.

In practical application, an individual object usually contains several different attribute variables, and the values of the attribute variables are required to be standardized, so that unified difference calculation is performed. The degree of importance of the attribute variables varies in most cases and is taken into account in the calculation of the degree of difference. The idea of variable value normalization is to convert all classification variables into two-state variables and all ordinal variables into interval variables. And if the attribute variables of the individual objects are standardized correspondingly, calculating the degree of difference between the objects. Taking m objects as an example, after the variable is normalized, the m objects are all composed of interval variables and two-state variables, and at the moment, the difference degree formulas of the a object and the b object in the m objects are calculated as shown in a formula 5, a formula 6 and a formula 7. Wherein the variables in the object a are x _a1,x_a2,x_a3…,x_an, the variables in the object b are x _b1,x_b2,x_b3…,x_bn, the number of the variables is n, the value of the kth variable of all the objects is x _1k,x_2k,x_3k…,x_mk,w_k, and when the k variable is an asymmetrically changed two-dimensional variable, and x _ak＝x_bk =0, the variable is generally considered to be unimportant, and thus the weight is 0.

In another possible implementation manner, project division is performed on project projects to be predicted to obtain a plurality of meta-projects, and then the method is implemented by inputting project data to be predicted into a database analysis model, obtaining a plurality of divided meta-projects, and then determining similar project groups matched with the project projects to be predicted.

As shown in fig. 3, fig. 3 shows an interface diagram representing the matching degree between a plurality of similar project projects in a database and project projects to be predicted, the interface diagram representing the matching degree between a plurality of similar project projects in the database and project projects to be predicted, in one example, a value interval is set, if the calculated value is closer to 1, the matching degree between two project projects is larger, and vice versa, and the matching degree is ranked according to the size.

As shown in fig. 3, a method for estimating construction cost of building engineering according to an embodiment of the present application is implemented according to the construction cost estimation system of building engineering and software thereof shown in fig. 1. In the unit engineering interface shown in fig. 3, the first column is an item column, the second column is a total cost amount, the third column is an item status, the fourth column is an engineering category, the fifth column is a professional class, the sixth column is a building area, the seventh column is an above-ground building area, the eighth column is an above-ground floor number, the ninth column is a standard floor height, the tenth column is a building height, the eleventh column is a base area, the twelfth column is a single cost, and the thirteenth column is a matching degree. The column names displayed by the interface are the item features that the input of the above mentioned model can include, but are limited to, and the modification of the feature names can be performed according to the actual parameter requirements and the data classification requirements.

And 103, performing index collection calculation to obtain meta-engineering quantity index data of each project engineering.

In this step, data collection, data processing, and meta-engineering quantity index calculation are involved. And collecting data, namely collecting the original data of all main indexes (such as reinforcing steel bars, cement, commercial concrete and the like). This may include information on the number, cost, etc. of use per project; and (3) data processing, namely cleaning and preprocessing the collected data according to the need, and ensuring the consistency and accuracy of the data. And mapping the same first 9 digits in the 12-digit codes corresponding to the minimum metering unit engineering quantity list items defined in the specifications into the same metadata engineering data item. Further standardizing according to the measurement units of the index so as to unify in subsequent calculation; and calculating the meta-engineering quantity index, namely calculating the meta-engineering quantity index of each project by using a corresponding calculation method to obtain specific numerical values of various indexes such as steel bars, cement and the like, and processing and calculating the data by using a corresponding calculation method and a corresponding formula to obtain meta-engineering quantity index data.

And 104, carrying out combined calculation on each piece of engineering quantity data contained in each engineering project and the corresponding engineering level matching degree to obtain each piece of engineering quantity index initial data contained in the engineering project to be predicted.

In one possible implementation, step 104 includes the following.

Content one, ensure the accuracy of the combined calculation. Before combined calculation, data consistency check is carried out to ensure that the data sources and calculation methods of each meta-engineering in the target meta-engineering group are consistent. If there is a difference in the unit of measure of the meta-engineering data, it is ensured that a normalization process is performed before the combination calculation to maintain the consistency of the data. Detecting and processing abnormal values in the target meta-engineering group, and avoiding the influence of abnormal data on a combined calculation result; content two, use project level matching degree algorithm. And (3) calculating the matching degree of each project in the similar project group by considering an algorithm for adopting the project hierarchical matching degree and combining the hierarchical structure of the project. When calculating the matching degree, considering the weight of project engineering characteristics, and enabling the matching degree calculation to be closer to the actual situation by modifying the weight; and thirdly, real-time feedback and verification of the matching degree. And during combined calculation, the matching degree information of each project is fed back in real time, so that a user can know the specific situation of the matching degree. Establishing a checking mechanism of the matching degree, ensuring that the calculation of the matching degree accords with the expectation, and avoiding errors caused by calculation errors; content IV, data updating strategy. And regularly checking historical engineering project data and algorithm for calculating the matching degree, and updating the data and the algorithm according to actual conditions so as to adapt to the changes of business and markets. When the problem of the data or algorithm is found, an automatic correction mechanism is designed, so that correction can be performed under the condition that the whole flow is not influenced; and fifthly, refining the combined calculation flow. The combined calculation flow is divided into a plurality of stages, and the metadata of different levels are gradually combined, so that the calculation complexity is reduced, and the calculation efficiency is improved. And the intermediate result of the combined calculation is visualized, so that a user can monitor and analyze the calculation process and result conveniently.

Therefore, when the initial data of the meta engineering quantity indexes of the projects of the similar project engineering group are calculated in a combined mode, the accuracy of calculation and the consistency of the data are ensured. Meanwhile, the matching degree information is fed back in real time by adopting an algorithm of the project engineering level matching degree, so that the matching process is more transparent and controllable.

And 105, inputting the metadata contained in each project and the initial data of each metadata engineering quantity index contained in the project to be predicted into a prediction difference adjustment model to obtain engineering quantity index correction coefficients corresponding to the metadata contained in the project to be predicted.

In one possible implementation, the following scheme is given for the construction of the predictive robust model. First, a model suitable for predicting the robust is selected. Regression models, such as linear regression, support vector regression, or more complex neural network models, may be considered, with particular choice depending on the complexity of the data and the model performance requirements; next, input features of the model are determined. In this scenario, the input of the model should include the respective meta-engineering data and the respective meta-engineering quantity index initial data contained in the project to be predicted. These data will be used to predict the engineering quantity index correction coefficients corresponding to each meta engineering data contained in the project to be predicted; further, the output of the model is explicitly defined. In this scenario, the output of the model is the engineering quantity index correction coefficient corresponding to each piece of meta engineering data; further, a training data set is prepared, wherein the training data set comprises each piece of meta-engineering data, each piece of meta-engineering quantity index initial data contained in project engineering to be predicted and corresponding actual correction coefficients. Ensuring diversity and coverage of the data set; further, the model is trained using the training data. According to the selected model type, adjusting parameters of the model to minimize an error between a prediction result and an actual correction coefficient; further, the trained model is evaluated using the validation dataset to check its performance on unseen data. The evaluation index may include a mean square error (Mean Squared Error, MSE), etc.; further, performing super-parameter tuning according to the evaluation result to improve the model performance; further, the trained model is applied to actual prediction. And inputting each piece of meta-engineering data and each piece of meta-engineering quantity index initial data of the project engineering to be predicted, and outputting engineering quantity index correction coefficients through the model.

In addition, the actual correction coefficients are collected periodically and compared with the model output. And updating and improving the model according to the comparison result so as to adapt to new data distribution. And finally, deploying the trained and evaluated prediction difference adjustment model into an actual system for real-time engineering quantity index correction.

And 106, generating a house construction project cost estimation report corresponding to the project to be predicted according to the project quantity index correction coefficient corresponding to each piece of meta project data contained in the project to be predicted and each piece of meta project quantity index initial data contained in the project to be predicted.

In addition, in this step, a multi-time, hierarchical group price process is used. For each meta-project, the modified project quantity index and the corresponding price (meta-project comprehensive unit price) are combined for multiple times of price grouping. And then summarizing step by step to form the final predicted price (project construction cost).

The report may include information such as a specific value of each main index, total cost, etc., and may be formatted and visualized as needed.

In one possible implementation, a scheme for generating a estimated construction cost report of the building is given as follows.

Firstly, the structure of a report is designed, so that the corrected engineering quantity index, the comprehensive unit price, the price combining process and the final project construction cost of each meta-project can be clearly displayed. The structure of the report should include, but is not limited to, the following information.

Project basic information: project name, project number, construction site, unilateral cost, project status, etc.

Meta engineering detailed information: and the engineering quantity index after correction of each element engineering, the comprehensive unit price and the like.

The price combining process comprises the following steps: the group price process of each meta-engineering comprises multiple times of group prices in a layered level.

Summarizing step by step: and step-by-step summarizing the price of each level of meta-engineering group.

Final predicted price: the final predicted price of the entire project.

Next, with respect to report contents, specific information to be included in the report is determined. In this scenario, the following report content may be considered.

Engineering quantity indexes after each element engineering correction: including quantity, unit of measure, corrected value, etc.

Meta engineering comprehensive unit price: the comprehensive unit price of each meta-engineering reflects the price change.

And (3) recording a price combining process: multiple, hierarchical group pricing processes for each meta-project are recorded for review and traceability.

Step-by-step summarizing information: the step-by-step summary information of the meta-engineering group prices of each step comprises the total price of each step.

Final project engineering cost: the final predicted price for the project may be the sum of all of the meta-engineering group prices.

Further, charts or other visualization means may be employed to enhance the readability of the report. For example, the corrected engineering quantity index of the different meta engineering is represented by a bar graph, and the trend of change in the group price process is represented by a line graph. In addition, a data quality cue is added to the report indicating the confidence level and confidence level of the correction coefficient and the group price process. This helps the user understand the accuracy of the predicted outcome. Finally, a report export function is provided to allow the user to save or share the generated report. Common export formats such as Excel, PDF are chosen.

In order to achieve more humanized design, user customization options can be added, so that a user can select specific information to be displayed or adjust the presentation mode of the report according to needs, and a data tracing function is added in the report, so that the user is allowed to trace back the source of each piece of meta-engineering data, the calculation mode of correction coefficients and the detailed record of the price combining process. In addition, the examination history of the report is recorded, including examination time, examination personnel and other information, so as to ensure the traceability and transparency of examination.

In the embodiment of the application, the visual elements are added by definitely defining the structure and the content of the report so as to improve the user experience, and meanwhile, the data quality prompt and tracing functions are added so as to ensure that the generated estimated construction cost report of the building engineering meets the user requirements, and the reliability and the operability are realized.

In summary, the construction cost prediction method of the building engineering can be carried on a construction cost prediction platform at a server end or a terminal, when the construction cost prediction method is used for predicting, the internal and external characteristic engineering data of the project to be predicted are obtained, the internal and external characteristic engineering data are input into a database analysis model to obtain similar project engineering groups which are accurately matched with the project engineering data, index collection calculation is performed to obtain meta-engineering quantity index data of each project engineering, combination calculation is performed on each meta-engineering data contained in each project engineering and the corresponding project engineering level matching degree of the meta-engineering quantity index data, initial data of each meta-engineering quantity index contained in the project to be predicted are obtained, and a prediction adjustment model is input together with the meta-engineering quantity index data of each project to obtain engineering quantity index correction coefficients corresponding to each meta-engineering data contained in the project to be predicted; the problems of low project cost data standardization degree, huge data screening processing capacity and difficult index establishment in the related technology can be solved by accurately matching similar project engineering groups and analyzing and calculating meta engineering data, and compared with the method relying on artificial experience, the project cost index can be rapidly and accurately generated and predicted by a computer, so that the management benefit and project investment management and control are improved, and the resource investment is saved and the social and economic benefits reach the standard.

Referring to fig. 4, a flowchart of a method for estimating construction cost of a building according to another exemplary embodiment of the present application is shown. The method comprises the following steps:

Step 401, acquiring historical engineering project data, and constructing a database analysis model according to the historical engineering project data.

In the background technology, the requirements of quick valuation, rechecking and calculation exist in the situation that design drawing data of a construction cost consultation enterprise is insufficient or depth requirements are not met in the early consultation stage of the project, so that accumulation and application of historical project data are particularly important. In order to solve the pain of the business scenario, the historical project data and the corresponding project information accumulated by the enterprise are required to be effectively utilized to form valuable cost index data. Historical engineering project data is collected, analyzed and entered.

Step 402, building meta-engineering data from historical engineering project data according to metering rules.

In one possible implementation, content of each historical engineering project is divided from historical engineering project data according to metering and pricing rules, so that meta engineering of different project contents is obtained.

Further, the engineering quantity list belonging to the same meta-engineering is automatically collected to obtain meta-engineering data corresponding to each meta-engineering, wherein the meta-engineering data are used for indicating meta-engineering quantity index data of the corresponding meta-engineering.

In a more detailed embodiment, step 402 may include the following.

And extracting the first content and the historical engineering project data.

The source of the data is considered, and the required information is extracted from the accumulated historical engineering project data, including engineering quantity lists, pricing rules, engineering project characteristics and the like. The integrity of the historical engineering project data is ensured, and the historical engineering project data comprises accurate engineering quantity data, project characteristics, cost and other information.

Content two, metering pricing rules map to meta-engineering.

And designing a mapping rule, namely mapping each item of content in the metering and pricing rule to the meta engineering, so as to ensure that each meta engineering can accurately reflect the corresponding metering and pricing rule. In the mapping process, accuracy is emphasized, the complexity and the diversity of rules are considered, and the mapping result is ensured to accord with the actual engineering situation.

Content three, meta-engineering classification and collection.

According to the metering and pricing rule, the historical engineering project is divided into different meta engineering, so that the rationality of division is ensured and the construction flow of the actual engineering is met. And introducing an automatic collection mechanism for engineering quantity items belonging to the same meta-engineering, so as to ensure the high efficiency and accuracy of collection.

And the content IV and the meta-engineering data indicate meta-engineering quantity indexes.

And correlating the meta-engineering data obtained by the aggregation with meta-engineering quantity index data to ensure that each meta-engineering has corresponding quantization indexes. And verifying the accuracy of the association, and ensuring that the meta-engineering data can accurately indicate the meta-engineering quantity index data of the corresponding meta-engineering.

By the scheme, the accuracy and the effectiveness of a series of steps of extracting and mapping the historical engineering project data to the meta engineering and dividing and automatically collecting the meta engineering are ensured. The construction method is favorable for constructing representative meta engineering, and provides a reliable basis for the subsequent project cost estimation.

And step 403, acquiring the internal and external characteristic engineering data of the project engineering to be predicted.

In this step, please refer to step 101, and the description of the embodiment of the present application is omitted here.

And step 404, inputting the internal and external feature engineering data of the project engineering to be predicted into a database analysis model, and determining similar project engineering groups matched with the internal and external feature engineering data of the project engineering to be predicted.

In this step, please refer to step 102, and the description of the embodiment of the present application is omitted here.

In one possible implementation, step 404 is followed by the following.

Acquiring main internal and external feature item names and corresponding values in internal and external feature engineering data of project engineering to be predicted, wherein the main internal and external feature item names are obtained according to preset selection; matching corresponding similar project engineering groups from the historical engineering project data according to the names and corresponding values of the main internal and external feature items and performing aggregation calculation; and feeding back meta-engineering data of each project in the similar project engineering group after calculation is completed.

Wherein, main internal and external feature item names: reinforcing steel bars, cement, commercial concrete, mortar, wood, stones, building blocks, bricks, doors, windows, heat insulation materials, waterproof coiled materials, waterproof coating materials, precast slabs and the like.

In another possible implementation, step 404 is followed by the following.

Adding custom collection labels to each meta-project in similar project engineering groups; judging and prompting the correctness of the added custom aggregation label according to the consistency requirement of the meta engineering and the corresponding metering unit; and updating and exporting the similar project engineering groups.

The self-defined collecting label can be steel bars, a structure, a two-structure, a PC component, a template, a masonry, an ALC wallboard, a floor, inner wall plastering, outer wall plastering, ceiling plastering, waterproof coiled materials and the like.

Further implementation of the content includes adding custom collection labels in similar project engineering groups, and the custom collection labels can be added into a 'one structure', 'two structures' or the like according to preset requirements, so as to more accurately distinguish the characteristics of meta engineering. And judging the correctness of the added custom collection label according to the consistency requirement of the meta engineering and the metering meta, and ensuring that the added label meets the engineering pricing specification and the actual engineering requirement. Similar project engineering groups are derived. And according to the calculated and updated similar project engineering groups, carrying out export for the use of subsequent steps. The exported project-like engineering group needs to keep real-time, and the data used in the prediction process is ensured to be up-to-date and accurate.

Through the scheme, the operation of adding and updating the custom tag is considered, so that the more flexible and accurate engineering feature matching requirement is met.

And step 405, performing index aggregation calculation to obtain meta-engineering quantity index data of each project engineering.

In this step, please refer to step 103, and the description of the embodiment of the present application is omitted here.

And 406, carrying out combined calculation on each piece of meta-engineering quantity data contained in each project and the corresponding project level matching degree to obtain initial data of each piece of meta-engineering quantity index contained in the project to be predicted.

In this step, please refer to step 104, and the description of the embodiment of the present application is omitted here.

Step 407, inputting the metadata included in each project and the initial data of each metadata engineering quantity index included in the project to be predicted into a prediction difference adjustment model to obtain the engineering quantity index correction coefficient corresponding to each metadata engineering data included in the project to be predicted.

In this step, please refer to step 105, and the description of the embodiment of the present application is omitted here.

Step 408, generating a estimated building construction cost report corresponding to the project to be predicted according to the engineering quantity index correction coefficient corresponding to each piece of meta engineering data contained in the project to be predicted and the initial data of each piece of meta engineering quantity index contained in the project to be predicted.

In this step, please refer to step 106, and the description of the embodiment of the present application is omitted here.

In summary, in the embodiment of the present application, the above description describes a technical process of predicting construction cost of a building, wherein a plurality of steps such as index aggregation, matching degree calculation, and prediction adjustment model application are involved. The method has the following technical effects:

By considering each meta-engineering quantity data contained in project engineering and the corresponding project engineering level matching degree, the meta-engineering quantity index is corrected by adopting a prediction difference adjustment model, and the prediction accuracy is expected to be improved. The training process of the model can be performed according to the actual sample capacity, which is helpful for capturing potential laws and trends.

A number of main indicators, such as steel bars, cement, etc., are considered, which makes the prediction more comprehensive. Through the combined calculation of indexes, the characteristics of meta engineering can be more comprehensively known, so that the prediction comprehensiveness is improved.

And models such as BP artificial neural network and the like are adopted, and model parameters are corrected through feedback, so that the models have learning and adaptation capabilities. Therefore, the model can be continuously optimized in the subsequent prediction, and the robustness of the prediction is improved.

The Z-score data standardization normalization processing is carried out before prediction, so that the dimension influence among different indexes can be eliminated, and the model has higher universality.

The weight influence of the meta engineering features is considered, namely the influence of the weight of different feature factors on the matching degree value is considered, and the matching degree calculation is more in line with the actual situation through weight correction, so that the matching accuracy is improved.

And generating a comprehensive report, wherein the finally generated estimated construction cost report of the building construction synthesizes the corrected meta-engineering quantity indexes, provides a comprehensive view of construction cost estimation, and is beneficial to decision making and planning.

The above embodiment provides a method for implementing a prediction method of building construction cost, and the following describes an algorithm formula related to main steps in combination with a specific algorithm.

In step 102, similar project engineering groups are described by taking a clustering algorithm as an example, but the method is not limited thereto, and the corresponding principle is described as follows.

Given a sample set Ω= { x ₁,x₂,…,x_m } of a similar project engineering group, the k-means clustering method first needs to determine the cluster center (mean vector) { c ₁,c₂,…,c_k } of each class, where c _i is the mean vector of the cluster S _i, and its calculation formula is:

where, |s _i | represents the number of samples in the i-th class. Then calculate the distance d _n(x_it-c_jt of each sample x _i＝(x_i1,x_i2,…,x_m) to the respective cluster center c _j＝(c_j1,c_j2,…,c_jn), taking the euclidean distance as an example, the calculation formula is as follows:

And dividing the sample into categories to which the cluster center nearest to the sample belongs according to the calculated distance to obtain cluster division S= { S ₁,S₂,…,S_k }. The index adopted for evaluating the quality of the classification result of the k-means algorithm is an average error E, and the calculation formula is as follows:

Wherein the method comprises the steps of The square of the L ₂ norm, representing the difference between sample x and mean vector c _i. The average error E measures how tightly the samples in a class are around the center of the cluster, with smaller E representing a higher degree of matching of the samples in the cluster. The smallest average error E corresponds to the best classification scheme.

The method is an optimization problem of different vector combinations, belongs to a non-deterministic problem of polynomial complexity, needs iterative solution, optimizes the problem in an iterative mode by a k-means algorithm, and comprises the following specific flows:

1. initializing, and randomly selecting k samples from a sample set as cluster centers given the number k of classes.

2. And in the dividing stage, the distances between each sample and k cluster centers are calculated one by one, and each sample is divided into categories to which the cluster center nearest to the sample belongs, so that k categories are formed.

3. In the updating stage, the cluster center of each class is recalculated, namely, the mean vector of all samples in each class is calculated as a new cluster center.

4. Judging the end of iteration, and when the distance between the current cluster center of each class and the cluster center of the last iteration is smaller than a set threshold value, ending the iteration and outputting a classification result; otherwise, repeating the second to fourth steps.

In the algorithm flow above, the initial cluster center is k samples selected randomly. In practical application, all samples can be randomly distributed into k categories, and each cluster center is calculated according to the randomly formed categories to serve as an initial cluster center.

For determining the number of clusters, there are cases where the problem of the initial number of clusters cannot be avoided. Through a large number of documents for reading, analyzing and processing actual problems, the initial clustering number is often the final clustering number, and the clustering number is proportional to the sample number. Therefore, the number of clusters needs to be preset by researchers during cluster analysis.

Hierarchical clustering algorithm, the key of the aggregate clustering algorithm is the calculation of inter-class distances, and the selection of a calculation method directly influences the clustering result, so that a proper algorithm is required to be selected according to different problems. There are various calculation methods for the inter-class distance, and there are commonly used minimum distance d _min(C_i,C_j), maximum distance d _max(C_i,C_j) and average distance d _avg(C_i,C_j).

Minimum distance d _min(C_i,C_j), for a given two classes C _i and C _j, minimum distance d _min(C_i,C_j) refers to the distance between the nearest samples in the two classes, namely:

d_min(C_i,C_j)＝min{dist(x,z)|x∈C_i,z∈C_j}

wherein dist is a generalized distance function marker. The aggregate clustering algorithm using the minimum distance as the inter-class distance calculation method can handle non-spherical clusters.

The aggregate clustering algorithm employing the maximum distance as the inter-class distance calculation method is not easily affected by noise and outliers, but is liable to destroy larger clusters and tends to form spherical clusters.

Average distance d _avg(C_i,C_j), which is the average of the distances between all samples in the two classes, namely:

The aggregation clustering algorithm using the average distance as the inter-class distance calculation method is also not easily affected by noise and outliers, but tends to form spherical clusters. In view of the type and distribution form of the sample data in the scheme, the clustering method is adopted to carry out aggregation clustering analysis.

In the process of completing the preliminary screening result of the similar project engineering group in step 102, the related internal and external characteristic engineering data can be further screened to further process the similar project engineering group screening.

Because different characteristic factor items of the internal and external characteristic engineering data have different influence on the overall engineering quantity index in practice, the weight of each characteristic factor item in clustering operation can be determined by an AHP (advanced high performance) analytic hierarchy process.

Firstly, constructing a hierarchical structure model, firstly determining an evaluation target, then defining an evaluation criterion, and then constructing a hierarchical structure model by the target, the evaluation criterion and an action scheme. In the model, the targets, the evaluation criteria and the action schemes are in different levels, and the relationship among the targets, the evaluation criteria and the action schemes is represented by line segments.

Secondly, the factors are compared with each other in pairs, after the hierarchical structure model is made, an evaluator gradually determines importance weights of the factors of each layer relative to the factors of the previous layer from the first criterion layer downwards according to own knowledge, experience and judgment, and the importance degree is represented by a scale a _ij by using a two-by-two comparison method.

When the scale a _ij is 1, defining the i factor as important as the j factor; when the scale a _ij is 3, defining the factor i to be slightly more important than the factor j; when the scale a _ij is 5, the definition of the i factor is more important than the j factor; when the scale a _ij is 7, defining the i factor is much more important than the j factor; when the scale a _ij is 9, the definition of the i factor is much more important than the j factor; when the scale a _ij is 2,4, 6 and 8, defining that the importance comparison result of the two factors i and j is in the middle of the importance comparison result; when the scale a _ij is the reciprocal, the comparison result of the importance of the two factors j and i is the reciprocal of the comparison result of the importance of the two factors i and j. Second, the judgment matrix used for the pairwise comparison method can specifically determine the importance weights of different factors of each layer through the scale. Furthermore, the feature vector of the judgment matrix is calculated, and after the judgment matrix is written out, the feature vector of the judgment matrix is required to be calculated.

Taking a 3×3 judgment matrix a as an example: step one, calculating the sum of the rows of A; second, calculate the average value of each row, because A has 3 columns, divide 3 in the averaging; and thirdly, normalizing, namely dividing each row by the sum of three rows to obtain the feature vector of A. Finally, consistency test of the judgment matrix is carried out, and the situation of inconsistent judgment can occur in pairwise comparison, so that whether judgment obtained by pairwise comparison in the judgment matrix A is consistent or not needs to be tested. And according to the feature vector calculated by the inconsistent judgment matrix, the person can draw an erroneous conclusion. A consistency check of the decision matrix is required. For this purpose, a consistency index CI is calculated, which is defined as:

In the above formula, n is the order of the judgment matrix, and λ _max is the maximum eigenvalue of the judgment matrix. Then, the randomness index CR is searched from the following table, the ratio CI/CR is calculated, when the CI/CR is smaller than 0.1, the consistency of the judgment matrix can be considered to meet the requirement, otherwise, the judgment is needed to be carried out again, and a new judgment matrix is written.

The method of calculating lambda _max is as follows.

First, a vector a _w,A_w =aw is calculated; in the second step, lambda _max is calculated,

The above-mentioned W _i is a corresponding judgment matrix (a _w)_i feature vector, the weight of different feature factors is uniform by default in the initial stage, and specific weight values ω of different feature factors applicable to the corresponding judgment matrix can be further determined for different engineering project performance types in the subsequent research stage.

In the explanation of step 102, the project projects are mentioned to be sequentially arranged according to the project level matching degree, a matching degree value lambda _xy (namely, cosine angle method) between the similar project group x and the internal and external feature project data y of the predicted project is calculated, and a calculation formula is shown as follows, wherein a _kx,a_ky refers to the kth data (namely, the kth internal and external feature project data of the project) in two data samples.

The Z-score data is normalized before the cost data is analyzed, and the calculation formula is shown as follows.

And then, after the specific weight pairs of different characteristic factors are considered to evaluate the matching degree value of similar meta-engineering, the matching degree value lambda is corrected according to the weight value omega corresponding to each meta-engineering characteristic item, and the calculation formula of the corrected matching degree value lambda' _{xy Repair tool} is shown as follows, wherein omega _k refers to the weight value corresponding to the kth internal and external characteristic item of the project engineering.

In one example, the predicted engineering/human/material/machine consumption index is calculated as an example.

In step 104, the calculation formula of the initial data of each meta-engineering quantity index included in the project to be predicted is:

P _x is the initial data of the x-th meta-engineering quantity index contained in project engineering to be predicted;

s _n is building area data of the nth similar project;

Lambda' _{Repair tool n} is the matching degree correction value of each item engineering level of each meta engineering data contained in each item engineering in the similar item engineering group;

P _n is the index data of each meta-engineering quantity corresponding to the nth project in the similar project group in the database;

and a is an ellipsis mark.

By the formula, the combination calculation can be carried out on each piece of meta-engineering data contained in each piece of project and the corresponding project level matching degree, so as to obtain each piece of meta-engineering quantity index initial data contained in the project to be predicted.

The predicted difference adjustment model in step 105 may also be compared with the P _x actual value to correct the correlation coefficient of the predicted difference adjustment model according to the feedback result, and by using the BP artificial neural network algorithm, the present application uses a two-layer neural network as an example, as shown in fig. 7, to illustrate a corresponding specific example including 3 meta-engineering quantity indexes, where R, S, T respectively represents different meta-engineering quantity indexes. In fig. 7, the two-layer neural network includes an input layer, an hidden layer, and an output layer, where the output layer corresponds to the initial data of each meta-engineering quantity index included in the project to be predicted. Where the hidden layer neuron number is x and the output layer neuron number is 1, in the example of fig. 7, there are 3 samples, the input layer and hidden layer have 3 neurons, and the numbers in the circles represent the numbers. And (3) modifying the calculated matching degree value lambda _xy to serve as the initial weight w _O (n) of each neuron of the hidden layer, introducing engineering cost indexes to calculate a modification coefficient delta w, namely adopting a gradient descent method when E _n is minimized, and obtaining a modification value delta w _ji (n) of the synaptic weight w _ji (n) after training.

As shown in fig. 5, the induced local area v _j (n) generated at the activation function input of neuron j is defined as follows.

Is the activation function, y _i (n) is the input signal, and the function signal y _j (n) at the output of neuron j is defined as follows.

In the figure, y _j (n) and d _j (n) are the actual output and the desired output of neuron j, respectively, and the error signal generated by the output of neuron j is defined as follows.

e_j(n)＝d_j(n)-y_j(n) (5-3)

Where d _j (n) is the j-th element of the desired corresponding vector d (n). The instantaneous error E _j (n) defining neuron j is defined as follows, where E _j (n) is the error signal generated by the output of neuron j.

The errors of all output layer neurons are added to get the total instantaneous error E _n of the whole network as defined below.

Wherein set C includes all neurons of the output layer. The BP algorithm minimizes E _n by iteratively modifying the weights, and applies a modified value Δw _ji (n) to the synaptic weight w _ji (n) using a gradient descent method that is proportional to the partial derivative δE (n)/δw _ji (n). This gradient is expressed as follows according to the differential chaining rule.

The partial derivative δE (n)/δw _ji (n) represents a sensitivity factor, and determines the search direction of w _ji in the weight space.

Differentiating e _j (n) on both sides of formula (4-5) to obtain:

differentiating y _j (n) on both sides of formula (4-3) to obtain:

Differentiating v _j (n) on both sides of formula (4-2) to obtain:

Finally, differentiating w _ji (n) on two sides of the formula (4-1) to obtain:

substituting formulas (4-7) to (4-10) into formula (4-6) to obtain:

The correction Δw _ji (n) applied to w _ji (n) is defined as:

where η is the learning rate of the error back propagation and the negative sign indicates the gradient decrease in the weight space.

Substituting formula (5-11) into formula (5-12) to obtain:

Δw_ji(n)＝ηδ_j(n)y_i(n) (5-13)

wherein δj (n) is a defined local gradient:

The derivation described above assumes that neuron j is at the output layer node, now due to the two-layer neural network, neuron j is an implied layer node, and when neuron j is at the implied layer, there is no specified expected response to the input neuron. The error signal of the hidden layer is recursively determined from the error signals of all neurons directly connected to the hidden layer neurons. Considering neuron j as a hidden layer node, the local gradient δj (n) of the hidden layer neuron is redefined according to equation (5-14) as:

As shown in fig. 6, the subscript j represents an hidden layer neuron, and the subscript k represents an output layer neuron.

The total instantaneous error E _n of the network is:

And (3) performing bias derivative on the function signal y _j (n) at two sides of the formula (5-17), so as to obtain:

where v _k (n) is the induced local area generated at the activation function input of neuron k (i.e., the input of neuron k), d _k (n) is the desired output of neuron k, and e _k (n) is the error signal generated by the output of neuron k.

For output layer neuron k, it induces a local area:

Solving the differential of the equation (5-21) to y _j (n) to obtain:

substituting formulas (5-20) and (5-22) into formulas (5-18) to obtain:

Substituting equation (5-23) into equation (5-16) yields a local gradient δj (n) of hidden layer neuron j as:

from the above, the correction value Δwji (n) of the synaptic weight connected to the neuron j by the neuron i is defined as follows, by substituting the combination formula (5-13) into the formula (5-24):

Δw_ji(n)＝ηδ_j(n)y_i(n) (5-25)

The activation function adopts a sigmoid function as follows; learning rate 0< eta <1; training according to the actual sample capacity (the training times are the number N of the sample meta-engineering in the library, and N is larger than or equal to 50 in general).

In step 105, regarding the calculation of the engineering quantity index correction coefficient, the calculated correction coefficient Δw _x of the x-th most similar element engineering cost index is taken and is the average value of the corrected values Δw _ji (N) of the trained N-time post-synaptic weights w _ji (N), namely:

as shown in fig. 7, after three training, the correction coefficient is the three-time average value. The predicted value calculation formula of the meta-engineering amount/manual consumption amount/material consumption amount/mechanical consumption amount after correction in step 106 is synthesized as follows:

Wherein, P _n is the index data of each meta-engineering quantity corresponding to the nth project in the similar project engineering group in the database;

s _x is building area data of the nth similar project;

Deltaw _n is an engineering quantity index correction coefficient corresponding to each piece of meta engineering data contained in the project engineering to be predicted;

P _Prediction is the meta-engineering quantity index estimated cost report data of the x-th meta-engineering of the project engineering to be predicted.

Referring to fig. 8, a block diagram of a construction project cost estimating apparatus according to an exemplary embodiment of the present application is shown, where the apparatus includes:

The data acquisition module 801 is configured to acquire internal and external feature engineering data of an engineering project to be predicted;

The data input module 802 is configured to input the internal and external feature engineering data of the project to be predicted into a database analysis model, determine a similar project engineering group matched with the internal and external feature engineering data of the project to be predicted, where the similar project engineering group is formed by at least one matched project engineering and each project engineering is sequentially arranged according to a project engineering level matching degree, and indicate the meta engineering quantity index data contained by the meta engineering data;

The aggregation calculation module 803 is configured to perform index aggregation calculation to obtain meta-engineering quantity index data of each engineering;

The combination calculation module 804 is configured to perform combination calculation on each piece of meta-engineering data included in each piece of engineering and a corresponding item engineering level matching degree thereof, so as to obtain each piece of meta-engineering quantity index initial data included in the to-be-predicted item engineering;

The prediction difference adjustment module 805 is configured to input each piece of meta-engineering data included in the project to be predicted and each piece of meta-engineering quantity index initial data included in the project to be predicted into a prediction difference adjustment model, so as to obtain an engineering quantity index correction coefficient corresponding to each piece of meta-engineering data included in the project to be predicted;

The cost estimation module 806 is configured to generate a building project cost estimation report corresponding to the project to be predicted according to the project quantity index correction coefficient corresponding to each piece of meta-project data included in the project to be predicted and each piece of meta-project quantity index initial data included in the project to be predicted.

Optionally, before the data obtaining module 801, the apparatus further includes:

The first acquisition module is used for acquiring historical engineering project data and constructing the database analysis model according to the historical engineering project data;

And the second acquisition module is used for constructing the meta-engineering data from the historical engineering project data according to the metering and pricing rule.

Optionally, the second obtaining module includes:

The first acquisition unit is used for dividing the content of each historical engineering project from the historical engineering project data according to the metering and pricing rule to obtain each meta-engineering;

and the second acquisition unit is used for automatically collecting engineering quantity list items belonging to the same meta-engineering to obtain meta-engineering data corresponding to each meta-engineering.

Optionally, after the data input module 802, the apparatus further includes:

And updating and exporting the similar project engineering groups.

Optionally, the apparatus further includes:

The first processing module is used for acquiring input project key feature information about project engineering data to be predicted;

the second processing module is configured to input the internal and external feature engineering data of the project to be predicted into a database analysis model, determine a similar project engineering group matched with the project engineering data to be predicted, and include:

The third processing module is used for carrying out standardized processing on the project characteristic key information and inputting the project characteristic key information into the database analysis model;

And the fourth processing module is used for matching similar project engineering groups corresponding to the project characteristic key information through the database analysis model.

Optionally, the prediction tuning module 805 includes:

The first difference adjustment unit is used for inputting the meta-engineering quantity index data contained in each project in the similar project engineering group and the initial data of each meta-engineering quantity index contained in the project to be predicted into a prediction difference adjustment model to obtain the matching degree of meta-engineering layers of the meta-engineering quantity index data contained in each project in the similar project engineering group;

and the second difference adjustment unit is used for carrying out arithmetic average on the matching degree of the meta-engineering layer level of each meta-engineering quantity index data contained in each engineering in the similar engineering group to obtain the engineering quantity index correction coefficient of each meta-engineering data contained in the engineering to be predicted.

The embodiment of the application also provides a computer readable storage medium, wherein at least one instruction, at least one section of program, code set or instruction set is stored in the storage medium, and the at least one instruction, the at least one section of program, the code set or instruction set is loaded and executed by a processor to realize the method for estimating the construction cost of the building, which is provided by each embodiment.

Alternatively, the computer-readable storage medium may include: read Only Memory (ROM), random access Memory (RAM, random Access Memory), solid state disk (SSD, solid STATE DRIVES), or optical disk, etc. The random access memory may include resistive random access memory (ReRAM, RESISTANCE RANDOM ACCESS MEMORY) and dynamic random access memory (DRAM, dynamic Random Access Memory), among others.

The foregoing embodiment numbers of the present application are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.

It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program for instructing relevant hardware, where the program may be stored in a computer readable storage medium, and the storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.

The foregoing description of the preferred embodiments of the present application is not intended to limit the application, but rather, the application is to be construed as limited to the appended claims.

Claims

1. The method for estimating the construction cost of the building construction is characterized by comprising the following steps:

2. The method for predicting construction cost of a building according to claim 1, wherein before the obtaining of the internal and external feature engineering data of the project to be predicted, the method further comprises:

3. The method of predicting construction cost of a building according to claim 2, wherein the constructing the meta-engineering data from the historical engineering project data according to a metering rule comprises:

4. The method for predicting construction cost of building projects according to claim 1, wherein after inputting the data of the internal and external feature projects of the project to be predicted into a database analysis model, the method further comprises:

5. The method for predicting construction cost of building projects according to claim 1, wherein after inputting the data of the internal and external feature projects of the project to be predicted into a database analysis model, the method further comprises:

And updating and exporting the similar project engineering groups.

6. The method for predicting construction cost of a building according to claim 1, further comprising:

7. The method for predicting construction cost of building engineering according to claim 1, wherein the inputting the metadata included in the project engineering and the initial data of the quantity index of each metadata included in the project engineering to be predicted into the prediction adjustment model to obtain the correction coefficient of the quantity index of the project corresponding to each metadata included in the project engineering to be predicted includes:

8. The method for predicting construction cost of building construction according to any one of claims 1 to 7, wherein the engineering quantity index correction coefficient is denoted as Δw _x, and the calculation formula is as follows:

9. A building construction cost prediction apparatus, the apparatus comprising:

10. A computer readable storage medium having stored thereon at least one instruction for execution by a processor to implement the method of construction project cost estimation of any one of claims 1 to 7.