KR100451940B1 - Data Analysis System and Method capable of Managing Customer Relations - Google Patents

Abstract

According to the present invention, if a user inputs a few conditions, the customer's information, which is the marketing target of the corporate management, is analyzed according to the characteristics of the customer using the K prototype, and the result can be easily recognized by the user. The present invention relates to a data analysis system having a customer management function that displays and outputs the data.

According to the present invention, the customer management of the data analysis system that visually displays and outputs the data in the database by segmenting, clustering and easily distinguishing the contents in the database using the K prototype based on the database and the element data to be analyzed by the user. A method, comprising: outputting a data input form for inputting data required for data analysis, temporarily storing data input through the data input form, normalizing data in the input database, and inputting the data Segmentation and clustering using the K prototype algorithm based on the collected data, and visually outputting the clustered results provides a customer management method of a data analysis system.

According to the present invention, it is possible to analyze the customer characteristics of the customer database and display the result so that the result can be easily recognized by a simple input operation of the user, and thus the user can easily establish a sales strategy based on this.

Description

Data Analysis System and Method Capable of Managing Customer Relations

The present invention relates to a data analysis system having a customer management function and a method thereof. More specifically, when a user inputs a few conditions, the information of customers who are the marketing targets of the corporate management is based on the K prototype. The present invention relates to a data analysis system and method having a customer management function that analyzes according to customer characteristics and displays and outputs the result in a graph that can be easily recognized by a user.

Recently, as the business management environment has changed from supplier-centered to consumer-centered, the success or failure of corporate management depends on how appropriately individual consumers' various needs are satisfied.

It is very important to collect and accumulate identifiable information about the target customers or purchase and marketing related information for decision making and marketing activities of corporate management. Information about individual customers' disposition or desire can be collected through analysis or survey of data such as transaction history, customer complaints, etc., and information useful for decision making can be derived through appropriate analysis and interpretation.

Management decision-making requires complex analysis involving large amounts of data processing and transformations, and is difficult to perform without a statistical knowledge base and expensive statistical analysis software.

In recent years, there has been a tendency to set up a Data Warehouse House or to build an integrated Data Mining environment, and the company's cost for building a data analysis environment has become higher in this context. Therefore, general marketers can easily and inexpensively analyze the difficulty of their customer information and prospect prediction without using expensive data mining system, as if using an office program such as Excel. The need for a new data analysis solution is readily available.

Due to the aforementioned needs, there are more than 100 products on the market under the name of Data Mining Tool, and their prices vary widely. Well-known products include "Statistical Analysis System (SAS)" Enterprise Miner "," Statistical Package for the Social Sciences "" Clementine ", IBM" Intelligent Miner "and Oracle" Darwin ".

Although the solutions vary slightly depending on the type, they provide a connection with a database, an analysis result derivation function, a filtering and data conversion function, and a macro function based on a statistical analysis technique, but have the following problems.

First, it is difficult to use without the help of a professional statistical analyst. The data analysis itself is based on statistics, making it difficult for general marketers or analysts without a statistical background to access and absolutely need the help of experts or external consultants. In terms of interface, there is no guide, so that it is difficult for a user to proceed without prior knowledge of data status and analytical process.

Second, it focuses only on functioning as an analysis tool, not on actual business. The analysis process and analysis results provided by the tool have to be analyzed and applied to the user's problem, which is insufficient to apply to the specific business problem. It is not easy to select an appropriate analysis model and algorithm to solve the issue according to the user's business issue, and there is a problem that a lot of trial and error is required in selecting and optimizing the analysis model and algorithm.

Third, it is a matter of price and construction period. In general, data mining tools are very large, costing only $ 100 million in purchase costs, and in addition to the purchase price of data mining tools themselves, consulting costs are additionally required, and engineers and consultants to build the tools are required for a long time. There is a problem that needs to be supported.

Finally, not only is it difficult to connect to external databases or data files, but the data loading process for users entering data directly is cumbersome and detailed, making it difficult for analysts without computer expertise to use. Since the data conversion and standardization process must be performed by combining complex functions requiring the judgment and insight of the user, there is a problem in that the quality of data analysis varies greatly depending on the user, thereby ensuring the reliability of the analysis results.

In order to solve the above problems, the present invention provides a customer management function that analyzes the contents of a database in which customer information is stored using a K prototype based on a condition input by a user and displays and outputs the results. The purpose is to provide a data analysis system having.

Another object of the present invention is to input a few conditions, based on the information of the customers of the marketing target of the company management based on the characteristics of the customer using the K prototype, the results can be easily distinguished The present invention provides a method of analyzing customer management data that displays and outputs a graph.

According to an object of the present invention, a data analysis system having a customer management function that analyzes the contents of a database in which customer information is stored using a K prototype based on a condition input by a user and displays and outputs the result. A data input thread for presenting a key input unit for inputting or selecting data, a data input form for inputting data, and transferring the input data to the central processing unit, and standardizing the input data. Data conversion thread responsible for converting function, DB matching thread responsible for matching data in the database to be analyzed to the central processing unit, Algorithm thread for performing the operation of the K prototype algorithm, by the algorithm thread Responsible for plotting computed results And an analysis thread, a data storage unit for temporarily storing data inputted through the key input unit or an algorithm-processed intermediate value, a monitor for visually outputting the execution status of the system or the analyzed result of the database, and the key. A center for driving the data input thread, the data conversion thread, and the algorithm thread based on the data input through the input unit to read and analyze the contents of the database to be analyzed, and output the analysis result to the monitor. Provided is a data analysis system having a customer management function comprising a processing device.

According to another object of the present invention, based on the database and the element data to be analyzed by the user, using the K prototype, the contents of the database can be segmented and clustered to visually display and output for easy discrimination. A method for analyzing management data, the method comprising: outputting a data input form for inputting data required for data analysis, temporarily storing data input through the data input form, and data in a database based on the input data Standardizing the data, segmenting and clustering using the K prototype algorithm based on the input data, and visually outputting the clustered results are provided. .

1 is a block diagram schematically showing the configuration of a data analysis system having a customer management function according to an embodiment of the present invention;

2 is a flowchart illustrating a customer management method of a data analysis system according to an exemplary embodiment of the present invention;

3 is a view showing a data input form for inputting data;

4 is a diagram showing the contents of a database to be analyzed by the central processing unit;

Figure 5a is a bar graph output of the results analyzed using the K prototype,

Figure 5b is a diagram showing the results of the analysis using the K prototype as a donut graph.

<Explanation of symbols for the main parts of the drawings>

10: key input unit 20: data input thread

21: Database Field 22: Cluster Field

23: Attribute list input field 24: Independent variable input field

25: independent variable selection button 26: dependent variable input field

27: dependent variable selection button 30: data conversion thread

40: DB Matching Thread 50: Algorithm Thread

60: result analysis thread 70: data storage

80: monitor 90: central processing unit (CPU)

Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings.

First, in order to help the understanding of the present invention, the concept of a data analysis system having a customer management function according to the present invention will be briefly described, and the terms used in the present invention will be briefly described.

The data analysis system having a customer management function according to the present invention is a program in the form of a solution, which is used for data analysis of a file name of a database to be analyzed by a user by executing a data analysis system program and the elements in the database. In addition to specifying independent and dependent variables, the number of clusters can be set arbitrarily. Here, the number of clusters is the number of homogeneous customer groups desired by the user, and the independent variable is a variable representing an attribute that is a standard for subdividing the contents of the database such as the job, address, and age of the customer, and the dependent variable is an independent variable. It is a variable that represents the attributes that you want to know more for each cluster through the segmentation used. For example, the late fees, delinquency rate, purchased items, and purchase amount of customers are shown.

As described above, based on the data input by the user, the data analysis system classifies and groups objects having similar properties using the K-Prototype algorithm to form a group, and forms a dependent variable and an independent variable for each group. By statistically processing the ratio for each variable factor and outputting the result on the screen, the user can grasp the customer's purchase tendency or characteristics at a glance. Here, the prototype is an algorithm made to perform a specific function or the like for a system, and an algorithm for performing a function according to the present invention is defined as a K prototype.

In addition, the data analysis system according to the present invention performs its function through each thread. When one program is referred to as a process, the thread refers to a unit of execution in one program. In other words, it refers to the flow of a program having independent program counters within a process. Therefore, the present invention represents each task as a thread, and multitasking is possible by allowing several such threads to be provided.

1 is a block diagram schematically showing the configuration of a data analysis system having a customer management function according to an embodiment of the present invention.

In FIG. 1, reference numeral 10 denotes a key input unit for selecting or inputting a database name containing customer related data to be analyzed, or selecting a number of clusters or independent and dependent variables, and reference numeral 20 denotes an analysis of the database. Data input thread for presenting a data input form for inputting the required element data and passing the input element data to the central processing unit 90 to be described later, reference numeral 30 standardizes the input data. Data conversion thread, which is in charge of converting into standard data for reference, reference numeral 40 is a DB matching thread, which is responsible for matching individual data in the database to be analyzed to apply to the system.

In addition, reference numeral 50 is an algorithm thread for performing the operation of the K prototype algorithm according to the data type, which is represented as Euclidean distance function for numerical data and as a percentage of the cluster prototype for categorical data. For example, the weight of categorical data for numerical data applies 1/2 of the standard deviation for numerical data.

In FIG. 1, reference numeral 60 denotes a function of selecting a result calculated by the algorithm thread 50 in order of a high ratio of factors of each attribute for each cluster, for example, plotting a bar graph or a donut graph. A result analysis thread, reference numeral 70 denotes a data storage unit for temporarily storing element data selected or inputted through the key input unit 10 or an intermediate value processed by an algorithm, and reference numeral 80 denotes an operation execution process or a database of a system. A monitor for visually outputting the analyzed result of the reference numeral 90 reads data included in the database to be analyzed, and based on the element data selected and input through the key input unit 10, the data input thread 20 or The data conversion thread 30 and the algorithm thread 50 are driven to analyze the data, and the analysis result is output to the monitor 80. A central processing unit (CPU) for control.

Next, the method of analyzing customer management data according to the present invention will be described with reference to the flowcharts shown in FIGS. 2A and 2B in relation to the operation of the data analysis system having the customer management function configured as described above.

When the user executes the operation of the system by using the keys provided in the key input unit 10 or by using the mouse button (step S2), the central processing unit that recognizes the key input for executing the system operation from the key input unit 10 ( 90 drives the data input thread 20 to output the data input form as shown in FIG. 3 to the monitor 80 (step S4).

The data entry form shown in FIG. 3 includes a database input field 21 for selectively inputting a database name or a file name of a database to be analyzed, a cluster input field 22 for inputting the number of clusters, and a list of field names of a database. Attribute list text box (23) to show, independent variable text box (24) to show the list of independent variables, independent variable selection button (25) to set as an element of the independent variable from the property list, and a list of dependent variables The dependent variable input field 26, and the dependent variable selection button 27 for setting the element of the dependent variable in an attribute list, etc. are comprised.

The user checks the data input form output to the monitor 80, and inputs data required by an input box using a plurality of keys provided in the key input unit 10.

That is, the user selects a database to be analyzed through the database input field 21 or selects a file name in the form of a file, for example, Access or Excel. When a database or file is selected and input through the database input field 21 (step S6), the central processing unit 90 drives the DB matching thread 40 to display the contents of the selected database as shown in FIG. 4. In addition, the contents are temporarily stored at a predetermined address of the data storage unit 70, and the attribute list corresponding to the field of the database is output to the attribute list input field 23 (step S8).

For example, in the case of the database content shown in FIG. 4, the job, address, age, etc. corresponding to the column name or the field name may be selected. The attribute list input field 23 shows a list of database attributes.

Subsequently, the user selects and inputs the desired number of clusters through the cluster input field 22, and the central processing unit 90 temporarily stores data about the input number of clusters in the data storage unit 70. In addition, in the data input form shown in FIG. 3 output to the monitor 80, the user designates a list to be used as an independent variable from the list shown in the attribute list input field 23, and clicks the independent variable selection button 25. To set the independent variable input field 24. Similarly, the list of attributes to be used as the dependent variable is designated, and the dependent variable selection button 27 is clicked to move to the dependent variable input box 26.

As shown in FIG. 4, the database to be analyzed can be seen that the corresponding data are disposed in each field corresponding to the attribute. Fields corresponding to attributes are set as text, number, date / time, etc. as the data type when designing the database, so the CPU 90 based on the set data type It is to recognize whether data belonging to each field is categorical data or numeric data. That is, when the central processing unit 90 drives the DB matching thread 40 to read the contents of the database to be analyzed, the occupation, the address, the purchased item, etc. among the fields shown in FIG. 4 are categorical data. The data belonging to the age or purchase amount field is recognized as numerical data.

When the data input other than the database, that is, the number of clusters, independent variables, dependent variables, etc. is completed by the above-described process (step S10), the central processing unit 90 stores the input data in the data storage unit 70. ) Is temporarily stored at a predetermined address (step S12).

Then, the central processing unit 90 drives the data conversion thread 30 to normalize the numeric data among the data read from the database to 0 and 1 (step S14). This standardization is achieved by dividing each data into the largest of the numerical data, where the largest value is 1 and the smaller data is converted into data between 0 and 1. This standardization reduces the deviation from categorical data by converting the numerical value to a relative value between 0 and 1 even if it has a large value such as 10,000 or 100,000.

Subsequently, the central processing unit 90 randomly sets clusters as many as the number of clusters input in the cluster input field 22 based on the data input through the data input form, and randomly sets a prototype that is the center of the cluster for each cluster. It is set (step S16).

The central processing unit 90 determines whether a numerical attribute exists in each of the clusters set (step S18). When the numerical attribute exists in the inspected cluster (step S20), the CPU 90 drives the algorithm thread 50. The K prototype algorithm as shown in Equation 1 is executed to calculate the numerical data as a Euclidean distance function (step S22).

In Equation 1, j is an independent variable or a dependent variable, i is a cluster, x is numeric data, q is a prototype in a cluster, l is a cluster number, and r is a symbolic symbol representing a numeric type. Also, Represents the value of the j th numeric attribute of the i th data, Denotes the prototype of the j th numeric attribute of the l th cluster. Therefore, the distance value between the i th data of the j th numeric attribute and the j th prototype of the l th cluster is obtained by Equation 1.

In step S20, as a result of checking whether there is a numeric attribute for each of the set clusters, if there is no numeric attribute in the cluster, all data read from the database is randomly assigned to each cluster (step S24). .

Next, the central processing unit 90 drives the algorithm thread 50 and calculates a ratio of arbitrary data in the cluster to which the categorical data belongs to the prototype as shown in Equation 2 (step S26).

In Equation 2, Represents the value of the j th categorical attribute of the i th data, Denotes the prototype of the j-th categorical attribute of the l-th cluster, and δ denotes a function for calculating the proportion of prototypes in the cluster. Also, Means Is an element of all the factors of the j th categorical attribute in the cluster. And p is in l cluster Represents the proportion of Denotes the number of data in the cluster, Denotes the number of categorical attributes.

Therefore, Equation 2 means that arbitrary data Accounted for in the cluster is the prototype for the whole (1). Subtract the ratio of to obtain it.

Subsequently, the central processing unit 90 obtains the homogeneous values of the data and the prototypes in each cluster as shown in Equation 3 (step S28).

In Equation 3, Represents the value of the j th numeric attribute of the i th data, Represents the prototype of the j th numeric attribute of the l th cluster, Denotes the weight of the categorical attribute with respect to the numeric attribute of the l-th cluster.

And, Represents the value of the j th categorical attribute of the i th data, Denotes the prototype of the j-th categorical attribute of the l-th cluster, and δ denotes the proportion of the prototype in the cluster.

remind Is the weight of the categorical data for the numeric data, and is obtained by applying 1/2 of the standard deviation for the numeric data.

That is, Equation 3 means that homogeneous values of data and prototypes in each cluster are obtained by adding numerical values to weighted categorical values.

The CPU 90 performs clustering between entities having similar properties by using a segmentation algorithm such as Equation 4, that is, a K prototype algorithm, based on the obtained values (step S30).

In Equation 4, Represents the value of the j th numeric attribute of the i th data, Represents the prototype of the j th numeric attribute of the l th cluster, Denotes the weight of the categorical attribute with respect to the numeric attribute of the l-th cluster.

And, Represents the value of the j th categorical attribute of the i th data, Denotes the prototype of the j-th categorical attribute of the l-th cluster, and δ denotes a function for calculating the proportion of prototypes in the cluster.

Also, Indicates whether i data belongs to l cluster, Indicates that the value among clusters from cluster I to cluster k is the minimum. That is, Equation 4 means that arbitrary data In this case, the sum of distances of all data including categorical data and numerical data is minimized for each cluster to obtain the cluster having the minimum value.

The central processing unit 90 is random data For clusters, we assign the cluster with the smallest value, and perform this operation on all data.

According to the algorithm of Equation 4, the central processing unit 90 that performs the clustering between entities having similar attributes in the cluster determines whether or not clustering by the K prototype is performed for all the clusters set (step S32). If not already performed for all clusters, the process returns to step S20 for the next cluster and performs clustering by the K prototype (step S34).

Subsequently, the central processing unit 90 which drives the algorithm thread 50 to perform clustering by the K prototype for each cluster drives the result analysis thread 60 to calculate the result as shown in Equation 5 and monitor ( 80, the result is displayed and output. In the case of a numerical attribute, the result is displayed as a bar graph as shown in FIG. 5A.

In Equation 5, " cluster i * independent variable j * factor k " belongs to cluster i and represents factor k which is independent variable j.

The result analysis thread 60 distinguishes the clustering result output to the monitor 80 by using a bar graph so that a user can easily compare the ratio in the cluster for each factor in the attribute with the ratio in the whole in the case of the numeric attribute. In the case of categorical attributes, the donut graph is used to display the ratio in the cluster for each factor in the attribute and the ratio in the whole to the outside.

That is, according to the embodiment, when a user inputs a database to be analyzed and the number of independent variables, dependent variables, and clusters to be used for data analysis, the analysis system uses the K protocol based on the data input by the user. Using the K-Prototype algorithm, clustering of entities with similar properties is achieved, and the ratios of dependent and independent variable factors are statistically processed for each group and the result is displayed on the screen.

It will be appreciated by those skilled in the art that the present invention is not limited to the above embodiments, and that various modifications and changes can be made without departing from the technical scope of the present invention.

As described above, according to the present invention, the user's characteristics of the customer database are analyzed and displayed and outputted so that the result can be easily recognized by a simple input operation of the user, so that the user can easily establish a sales strategy based on this. It works.

Claims (13)

In the customer management data analysis method for visually displaying and outputting the data in the database by segmenting, clustering and easily distinguishing the contents of the database using a K prototype algorithm based on the database and the element data to be analyzed by the user. (a) outputting a data input form to a monitor to receive raw data required for data analysis from a key input unit; (b) temporarily storing the raw data input through the data input form in a data storage unit; (c) standardizing data in a database based on the row data temporarily stored in step (b) to generate normalized data in a central processing unit; (d) segmenting and clustering data in a database using a K prototype algorithm based on the standardized data to generate clustered data at the central processing unit; (e) outputting the clustered data to the monitor using a chart Wherein, the K prototype algorithm minimizes the sum of distances of all data, including categorical data and numerical data, for each cluster, for each cluster, and selects the cluster having the minimum value for the random data. Customer management data analysis method characterized in that the algorithm determined by the cluster. The method of claim 1, The data required for analyzing the data is a database, cluster number, independent variable, dependent variable, characterized in that the customer management data analysis method. The method of claim 1, Standardizing data in the input database is characterized in that the central processing unit converts the numeric data into data between 0 and 1. The method of claim 1, Segmentation and clustering using the K prototype algorithm is calculated by the central processing unit as a Euclidean distance function for numerical data, and for categorical data as a percentage of prototypes in a cluster. How to analyze customer care data. The method of claim 1, The K prototype algorithm , Where Is the value of the j th numeric attribute of the i th data, Is the prototype of the j th numeric attribute of the l th cluster, Is the weight of the categorical attribute for the numeric attribute of the l th cluster, Is the value of the j th categorical attribute of the i th data, Is the prototype of the j-th categorical attribute of the l-th cluster, δ is a function that calculates the proportion of prototypes in the cluster, Whether i data belongs to l cluster, Is a cluster from the I cluster to the k cluster, the value of the customer management method of the data analysis system, characterized in that the minimum. The method of claim 4, wherein The Euclidean distance function is Where j is an independent or dependent variable, i is a cluster, x is numeric data, q is a prototype within the cluster, l is a cluster number, r is a symbolic symbol representing a numeric type, Is the value of the j th numeric attribute of the i th data, Represents a prototype of the j th numeric attribute of the l th cluster, and the meaning is a data analysis system characterized by obtaining a distance value between the i th data of the j th numeric attribute and the j th prototype of the l th cluster Customer care method. The method of claim 4, wherein The proportion of prototypes in the cluster , Where Is the value of the j th numeric attribute of the i th data, Is the prototype of the j th numeric attribute of the l th cluster, Is the weight of the categorical attribute for the numeric attribute of the l th cluster, Is the value of the j th categorical attribute of the i th data, Is the prototype of the j-th categorical attribute of the l-th cluster, δ is a function that calculates the proportion of prototypes in the cluster, Denotes the weight of the categorical data with respect to the numeric data, and the weight applies 1/2 of the standard deviation for the numeric data. A method for managing a customer of a data analysis system, characterized by indicating that the numerical value is added to the categorical value to which the weight is applied. In the data analysis system having a customer management function that analyzes the contents of the database in which the customer information is stored using the K prototype based on the conditions entered by the user and displays the result. A key input unit for inputting or selecting data; A data entry thread presenting a data entry form for data entry and passing the input data to the central processing unit; A data conversion thread responsible for converting data in the database into standardized data, DB matching thread that is responsible for matching the data in the database to be analyzed to the central processing unit, An algorithm thread for executing a K prototype algorithm on data normalized based on the input data input from the key input unit, A result analysis thread in charge of the function of tabulating the results computed by the algorithm thread, A data storage unit for temporarily storing data inputted through the key input unit or intermediate values of algorithms processed; A monitor that visually outputs the operation execution status of the data analysis system or the analyzed result of the database; Based on the data input through the key input unit, the data input thread, the data conversion thread, and the algorithm thread are driven to read and analyze the contents of the database to be analyzed, and output the analysis result to the monitor. Central processing unit Wherein, the K prototype algorithm minimizes the sum of distances of all data, including categorical data and numerical data, for each cluster, for each cluster, and selects the cluster having the minimum value for the random data. A data analysis system having a customer management function, which is an algorithm determined by a cluster. The method of claim 8, The algorithm thread calculates the Euclidean distance function if the target of the operation is numeric data and calculates the ratio of the prototype of the cluster in the case of categorical data. A data analysis system with a customer care function, characterized by applying one half of the standard deviation for the type data. The method of claim 8, The data input form output by the data input thread includes a database input box, a cluster input box, an independent variable input box, a dependent variable input box, an attribute list input box, an independent variable selection button, and a dependent variable selection button. Data analysis system. The method of claim 8, The data conversion thread is a data analysis system having a customer management function, characterized in that to standardize the numeric data of the data read from the database between 0 and 1. The method of claim 8, The result analysis thread outputs a bar graph when the analysis result is a numerical type, and outputs the result as a donut graph when categorical. The method of claim 8, The algorithm thread computes the Euclidean distance function for numeric data and calculates the ratio of prototypes in the cluster for categorical data. , Where Is the value of the j th numeric attribute of the i th data, Is the prototype of the j th numeric attribute of the l th cluster, Is the weight of the categorical attribute for the numeric attribute of the l th cluster, Is the value of the j th categorical attribute of the i th data, Is the prototype of the j-th categorical attribute of the l-th cluster, δ is a function that calculates the proportion of prototypes in the cluster, Whether i data belongs to l cluster, Is a data analysis system with a customer management function, characterized in that the value of the cluster from cluster I to cluster k is the minimum.