CN109299994B - Recommendation method, device, equipment and readable storage medium - Google Patents
Recommendation method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN109299994B CN109299994B CN201810843390.7A CN201810843390A CN109299994B CN 109299994 B CN109299994 B CN 109299994B CN 201810843390 A CN201810843390 A CN 201810843390A CN 109299994 B CN109299994 B CN 109299994B
- Authority
- CN
- China
- Prior art keywords
- user
- data
- application scene
- preference
- historical
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Marketing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a user preference recommendation method, a device, equipment and a readable storage medium, which are used for extracting user characteristic data under each application scene from historical data; collecting characteristic vectors of historical comment data in each application scene; calculating similarity parameters aiming at each application scene by adopting the characteristic vectors; generating preference degrees of each application scene by adopting the user characteristic data and the similarity parameters; and recommending based on the preference degrees of the application scenes. The problem that recommendation can not be performed according to the demand preference of the user under the extensive scene in the prior art is solved.
Description
Technical Field
The embodiment of the invention relates to the technical field of electronics, in particular to a user preference recommendation method, device, equipment and a readable storage medium.
Background
The popularization of social applications and shopping applications and the generation of a large amount of user data, wherein the user preference statistics according to the user data is a commonly used method for various social applications and shopping applications so as to extract the characteristic interests of users and further achieve the purpose of providing personalized services for the users.
In the prior art, the preference of a user to a merchant is mined mainly through characteristics of user characteristics and merchant dimensions, for example, by processing microblog text data of the user, words with high occurrence frequency in the text are determined as microblog candidate topics which serve as expressions of user interest characteristics, and microblogs with similar topics are recommended to the user.
However, the prior art cannot recommend the demand preference of the user in a wide range of scenes.
Disclosure of Invention
The invention provides a user preference recommendation method, which aims to solve the problem that recommendation can not be carried out according to the requirement preference of a user under a broad scene in the prior art.
According to a first aspect of the present invention, there is provided a user preference recommendation method, the method comprising:
extracting user characteristic data under each application scene from historical data;
collecting characteristic vectors of historical comment data in each application scene;
calculating similarity parameters aiming at each application scene by adopting the characteristic vectors;
generating preference degrees of each application scene by adopting the user characteristic data and the similarity parameters;
and recommending based on the preference degrees of the application scenes.
According to a second aspect of the present invention, there is provided a user preference recommending apparatus, the apparatus comprising:
the user characteristic data extraction module is used for extracting user characteristic data under each application scene from historical data;
the comment feature vector acquisition module is used for acquiring feature vectors of historical comment data in each application scene;
a similarity parameter obtaining module for calculating similarity parameters for each application scene by using the feature vectors;
the application scene preference degree generating module is used for generating preference degrees of all application scenes by adopting the user characteristic data and the similarity parameters;
and the preference recommending module is used for recommending based on the preference degrees of the application scenes.
According to a third aspect of the invention, there is provided an apparatus comprising:
a processor, a memory and a computer program stored on the memory and executable on the processor, characterized in that the processor implements a user preference recommendation method as described in the foregoing when executing the program.
According to a fourth aspect of the present invention, there is provided a readable storage medium, wherein instructions, when executed by a processor of an electronic device, enable the electronic device to perform the aforementioned user preference recommendation method.
According to the user preference recommendation method, device and equipment and the readable storage medium provided by the embodiment of the invention, user characteristic data under each application scene is extracted from historical data; collecting characteristic vectors of historical comment data in each application scene; calculating similarity parameters aiming at each application scene by adopting the characteristic vectors; generating preference degrees of each application scene by adopting the user characteristic data and the similarity parameters; and recommending based on the preference degrees of the application scenes. According to the proposal, the purpose of recommending the preference application scenes to the user is realized by aiming at the preference degree of each application scene and the similarity parameter of each application scene.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments of the present invention will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive labor.
FIG. 1 is a flowchart illustrating steps of a method for recommending user preferences according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating steps of a method for recommending user preferences according to an embodiment of the present invention;
FIG. 2A is a schematic diagram of an implementation structure of a user preference recommendation method according to an embodiment of the present invention;
FIG. 2B is a graphical illustration of user historical behavioral preferences plus time decay provided by an embodiment of the invention;
FIG. 2C is a schematic diagram of a user scene preference sparse matrix provided by an embodiment of the present invention;
fig. 2D is a schematic diagram illustrating a comparison of personalized scene drainage effect statistics provided by an embodiment of the present invention;
FIG. 3 is a block diagram of a user preference recommending apparatus according to an embodiment of the present invention;
fig. 4 is a block diagram of a user preference recommending apparatus according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The following first introduces terms involved in embodiments of the present invention:
an LDA (latent Dirichlet distribution model) is a document theme generation model, comprises three-layer structures of words, themes and documents, and is mainly used for processing large-scale text clustering and extracting themes by using a statistical sampling method.
Word2vec, a model for converting words into vectors understandable by computer, is mainly used for processing the problem of vector representation of short texts
The collaborative recommendation is a method for recommending similar purchased articles by finding similar users, and in the proposal, an ALS (alternating least squares) method is used for filling a matrix of users and subjects, so that the problem of sparsity of the user matrix is solved.
Example one
Referring to fig. 1, a flowchart of steps of a user preference recommendation method is shown, which includes the following specific steps:
in the embodiment of the invention, in each user historical data recorded in an application background database, a feature extraction algorithm is used, for example, LDA is used for extracting features of comment data issued by a user aiming at each application scene, word2vec is used for extracting features of historical search records of the user aiming at each application scene, and then, according to basic information of the user, such as gender, age, occupation, marital state and the like, a user historical behavior log and collaborative recommendation are used for filling and making up user sparse preference, and user feature data corresponding to each application scene are generated.
And step 102, collecting characteristic vectors of historical comment data in each application scene.
In the embodiment of the invention, the application scene generally refers to a merchant scene, a unified text is generated after historical comment data in each application scene is collected, and then a text feature algorithm (doc2vec) is utilized to obtain the feature vector of the unified comment text of each application scene.
103, calculating similarity parameters aiming at each application scene by adopting the characteristic vectors;
in the embodiment of the invention, the comment text feature vector corresponding to each application scene is used as the similarity of each application scene, and the behavior preference of a user and a merchant is converted into the label of the preference of the user scene, namely the similarity parameter of the application scene.
And 104, generating preference degrees of each application scene by adopting the user characteristic data and the similarity parameters.
In the embodiment of the present invention, the user feature data obtained in step 101 and the similarity parameter obtained in step 103 are input into a non-linear classification model, such as a support vector machine, and the output value is the preference of the user for each application scenario.
And 105, recommending based on the preference degrees of the application scenes.
In the embodiment of the invention, different application scenes are recommended to the user according to the preference degree of the application scenes of the user. The recommended application scenes may be displayed in a sequential manner according to the preference of the user, or only the application scenes with the strongest preference may be displayed, which is not limited in the present invention.
In summary, according to the user preference recommendation method provided by the embodiment of the present invention, user feature data in each application scenario is extracted from historical data; collecting characteristic vectors of historical comment data in each application scene; calculating similarity parameters aiming at each application scene by adopting the characteristic vectors; and generating the preference degree of each application scene by adopting the user characteristic data and the similarity parameter and recommending the preference degree based on the preference degree of each application scene. According to the recommendation method and the recommendation device, the purpose of more finely recommending the preference application scenes to the user is achieved through the preference degree of the user for each application scene and the similarity parameter of each application scene.
Example two
Referring to fig. 2, a flowchart of steps of a user preference recommendation method is shown, which includes the following specific steps:
step 201, extracting comment subject distribution characteristic vectors of users by performing LDA processing on history comment data of the users in various application scenes in the history data.
In the embodiment of the present invention, as shown in fig. 2A, a first part describes a process of user data preparation, where 1 is to perform feature extraction by using an LDA algorithm on historical comment data of a user for each application scenario in historical data stored in an application background.
The lda (latent Dirichlet allocation) is a document topic generation model, which is also called a three-layer bayesian probability model, and includes three layers of structures, i.e., words, topics, and documents. The generative model is a process of "selecting a topic with a certain probability and selecting a word from the topic with a certain probability" if each word of an article is considered. Document-to-topic follows a polynomial distribution, and topic-to-word follows a polynomial distribution. LDA is an unsupervised machine learning technique that can be used to identify underlying topic information in large-scale document collections (document collections) or corpora (corpus). It adopts bag of words (bag of words) method, which treats each document as a word frequency vector, thereby converting text information into digital information easy to model. Each document represents a probability distribution of topics, and each topic represents a probability distribution of words.
Specifically, after the historical comment data of the user for each application scene is processed by the LDA, a comment word frequency vector of each user for each application scene is obtained, and the word frequency vector maps the preference of the user for each application scene to a certain extent.
In the embodiment of the present invention, as shown in the first part of fig. 2A, 2 is to obtain a search word vector by obtaining historical search data of a user in the historical data, taking all historical search data of the user as processing objects, and using a word2vec algorithm.
Wherein, word2vec is also called word templates, Chinese name "word Vector", which is used to convert words in natural language into Dense Vector (Dense Vector) that can be understood by computer. Even if the corpus is not sufficient, the obtained word vector matrix is still not sparse, and the relationship among the vectors can be shown.
Specifically, when a historical search word of a user is used as a corpus, a word2vec is used for processing to obtain a search word vector, a word2vec model is trained by using user comments, and the method is mainly used for processing the relevance of short texts; the method comprises the following steps of segmenting words of historical search words (query) of a user, finding word vectors through word2vec respectively, and aggregating the word vectors into an overall user query vector, wherein the mathematical expression is as follows:
query 1: vector M (a)1,a2,a3,…aN)
Query 2: vector N (b)1,b2,b3,…bN)
Query 3: vector Q (c)1,c2,c3,…cN)
General assembly
The method can enhance the influence of different queries in different dimensions, such as 'spicy hot pot', and the value of each query in a certain dimension representing the spicy is high, so that the user preference for the spicy is enhanced by adding the above methods.
And 203, combining the user basic information in the historical data, the comment subject distribution characteristic vector of the user and the user search word vector to generate user characteristic data of the user for each application scene.
In the embodiment of the present invention, as shown in fig. 2A, the user basic information in 3, for example, information such as gender, age, occupation, marital status, love status, hometown, and the like, is combined with the comment topic distribution feature vector of the user and the user search word vector obtained in step 201 and 202, and the above data are integrated into an overall user feature vector for each application scenario, the integrated data may be stored in a matrix form and processed in the next step, and of course, storage manners of the integrated data may be different for different processing methods or models, which is not limited in this embodiment of the present invention.
Preferably, the method further comprises the following steps:
and step S1, acquiring the strengthened collaborative recommendation data of other users in each application scene through the marked high-credibility user data in the historical data.
In the embodiment of the present invention, as shown in 5 in the first part of fig. 2A, a collaborative recommendation method is used to fill a sparse matrix of user preference for topics (as shown in fig. 2C), where U is a user and I is a topic, and in order to solve the problem of sparsity of users for topics, the ALS method adopted here is similar to other general collaborative recommendation methods, and the main principle is to recommend similar preferences through a plurality of similar users.
Preferably, step S1 includes sub-steps A1-A5;
and a sub-step A1, obtaining marked high-credibility user data under each application scene in the historical data.
In the embodiment of the invention, some users can be marked as high-credibility users aiming at each application scene in the historical data, for example, the users who often generate user behaviors under the application scene and have good reputation.
And a substep a2, obtaining the preference degree of the high-credibility user for each application scenario through the high-credibility user data.
In the embodiment of the invention, the preference degree of the high-reliability user to each application scene can be obtained by processing the data of the high-reliability user, for example, processing all comment data of the high-reliability user.
And a sub-step A3, extracting the historical behavior log of the high-credible user from the high-credible user data.
In the embodiment of the invention, the historical behavior logs of the high-reliability users are extracted, and the behavior vectors of the high-reliability users are further obtained.
And a substep A4, generating behavior vector characteristics of the high-reliability user in each application scene according to the historical behavior log of the high-reliability user by adopting an alternating least squares algorithm.
In the embodiment of the invention, the Alternating Least Square (ALS) algorithm is utilized to obtain the characteristic vectors of the behaviors of all the high-credibility users, if the high-credibility users are A, B and C, the behaviors of the high-credibility users are otherThe vectors are vecA，vecB，vecC。
And a substep A5, obtaining the preference degrees of other users in the historical data to each application scene according to the preference degrees and the behavior vector characteristics.
In the embodiment of the invention, the specific practice method of the reinforced behavior cooperation is as follows:
given that the preference degrees of the scenes corresponding to the high-credibility users A, B and C of a certain scene are prefA， prefB,prefcTheir ALS behavior vector is vecA，vecB，vecCThen, the preference of other users (such as user N) to the scene is calculated as:
therefore, the preference degrees of other users in the historical data for each application scene can be obtained.
Preferably, the method further comprises the following steps:
and step S2, combining the user basic information in the historical data and the strengthened collaborative recommendation data to generate user characteristic data of the user for each application scene.
In the embodiment of the present invention, as shown in fig. 2A, the user basic information in 3, for example, information such as gender, age, occupation, marital status, love status, hometown, and the like, is combined with the comment topic distribution feature vector of the user, the user search word vector, the user historical behavior time preference feature, and the enhanced collaborative recommendation data obtained in step 201 and step S1, and the above data are integrated into feature data of the user for each application scenario, where the integrated data may be stored in a matrix form and processed in the next step, and of course, storage manners of the integrated data may be different for different processing methods or models, which is not limited in this embodiment of the present invention.
And step 204, acquiring the user active time in each application scene.
In the embodiment of the invention, according to the historical user behavior log recorded by the application background, the active time period of the user for each application scene can be extracted, for example, the stay time of the user A in the business scene B is longest between 8-10 points in the morning every day. The application time may be extracted in different statistical manners, or the user active time may be obtained in combination with multiple statistical manners, which is not limited in the embodiment of the present invention.
Preferably, step 204 further comprises: sub-step B1-sub-step B2;
and a sub-step B1 of obtaining a user historical behavior log in the historical data.
In the embodiment of the present invention, as shown in the second part shown in fig. 2A, a user historical behavior log is extracted from historical data stored in an application background, and generally, the log stores the online time of the user and the user behavior in the online time.
And a substep B2, acquiring the average active period of the user, the active time period of the user and the active time of the user in each application scene according to the historical behavior log of the user.
In the embodiment of the present invention, the average user activity period, the user activity time period, and the activity time of the user in each application scenario in the historical user behavior log may be extracted as main parameters for determining the user activity.
Preferably, the method further comprises the following steps:
v1, combining the historical behavior log of the user in the historical data with the time weight to obtain the time preference characteristics of the historical behavior of the user;
in the embodiment of the present invention, as shown in the first part of fig. 2A, where 4 is the weight of the historical behavior characteristics of the user, the influence of the historical behavior of the user on the user preference over time is simulated.
As shown in fig. 2B, fusing user history (within T days) click, order placement, purchase, collection, and behavior review, and performing weight summation, the calculation formula is as follows:
wherein t isiIs the current date, thistoryAs the date on which the user's action occurred, ti-thistoryThe influence factor is smaller when the date of the user behavior is longer than the current date, and the influence factor is smaller when the time is longer, so that the user preference is influenced to be smaller when the user historical behavior is simulated along with the time.
And V2, generating user characteristic data of the user for each application scene by combining the user basic information in the historical data and the time preference characteristics of the user historical behaviors.
In the embodiment of the present invention, as shown in fig. 2A, the user basic information in 3, for example, information such as gender, age, occupation, marital status, love status, hometown, and the like, is integrated by combining the historical behavior time preference characteristics of the user obtained in step V1 into an integrated user feature vector for each application scenario, the integrated data may be stored in a matrix form and processed in the next step, and of course, storage manners of the integrated data may be different for different processing methods or models, which is not limited in this embodiment of the present invention.
This step is the same as step 102 and will not be described in detail here.
And step 206, calculating similarity parameters aiming at each application scene by using the feature vectors.
This step is the same as step 103 and will not be described in detail here.
And step 207, generating the preference degrees of the application scenes by using a preset nonlinear classification model and a sequencing algorithm according to the user characteristic data and the similarity parameters.
In the embodiment of the present invention, as shown in the third part shown in fig. 2A, the prediction model ranks user preference scenes, in this application, a support vector machine algorithm is used, user feature data generated in the first part shown in fig. 2A and a user scene similarity parameter generated in the third part are input to the support vector machine model, and preference ranks, that is, preference degrees, of the user for each application scene are obtained.
Among them, Support-Vector Machines (Support-Vector Machines) are mainly used for classification and regression analysis. Given a set of training samples, each label belongs to two classes, an SVM training algorithm builds a model, assigns new instances to one class or other classes, and makes them non-probabilistic binary linear classification. An example of an SVM model, such as a point in space, is mapped such that examples of the different classes are represented by a distinct gap that is divided as widely as possible. The new embodiments then map into the same space and are predicted to belong to a category based on their falling on the gap side. Most importantly, the input data can be implicitly mapped into a high-dimensional feature space to effectively perform nonlinear classification.
Specifically, when the user characteristic data and the similarity parameter are input, the similarity classification of the user for each application scene can be obtained, then the application scenes are ranked according to the similarity by using a ranking algorithm, and finally the preference of the user for each application scene is obtained.
Preferably, recommendation is performed based on the user active time and the corresponding preference degree in each application scene
In the embodiment of the present invention, for the user active time of each application scenario obtained in step 204, when it is detected that the user is online at the corresponding active time, different application scenarios are recommended to the user according to the preference of the application scenario of the user. The recommended application scenes may be displayed in a sequential manner according to the preference of the user, or only the application scenes with the strongest preference may be displayed, which is not limited in the present invention.
And step 208, determining at least one application scene according to the search words currently input by the user in the average user active period, the user active time period and the active time of the user in the application scene.
In the embodiment of the invention, after the user activity is obtained, in the user activity time period, namely the average user activity period, the user activity time period and the active time of the user in the application scene, when the user is online and a search keyword is input, at least one application scene is determined according to the application scene corresponding to the search word of the user. For example, if the user inputs "spicy, hot pot", it can be determined that the application scenario belongs to the "food category", wherein the application scenarios corresponding to the hot pot can be identified as the merchant scenario.
In the embodiment of the present invention, the application scenes are sequentially displayed to the user according to the application scene preference ranking obtained in step 208. The application display interface may display a plurality of application scenes according to user settings, or may display an application scene with the highest preference, which is not limited in the embodiment of the present invention.
Specifically, by performing preference recommendation to the user, the user agent drainage realizes a significant increase, as shown in fig. 2D, the user traffic increase tutoring for each personalized channel is significantly improved.
In summary, according to the user preference recommendation method provided by the embodiment of the present invention, the historical comment data of the user in each application scene in the historical data is subjected to LDA processing, so as to extract the comment subject distribution feature vector of the user; obtaining a user search word vector by performing word2vec processing on historical search data of a user in the historical data; combining the historical behavior log of the user in the historical data with the time weight to obtain the time preference characteristics of the historical behavior of the user; acquiring enhanced collaborative recommendation data of other users in each application scene through marked high-credibility user data in the historical data; combining user basic information in historical data with the enhanced collaborative recommendation data to generate user characteristic data of the user for each application scene; collecting characteristic vectors of historical comment data in each application scene; calculating similarity parameters aiming at each application scene by adopting the characteristic vectors; generating preference degrees of each application scene by adopting the user characteristic data and the similarity parameters; acquiring user active time in each application scene; and recommending based on the preference degrees of the application scenes. According to the proposal, the purpose of recommending the preference application scenes to the user aiming at the active time of the user is more refined through the activity of the user, the preference of the user to each application scene and the similarity parameter of each application scene is realized.
EXAMPLE III
Referring to fig. 3, a block diagram of a structure of a user preference recommending apparatus is shown, which is as follows:
a user characteristic data extraction module 301, configured to extract user characteristic data in each application scenario from historical data;
a comment feature vector acquisition module 302, configured to acquire feature vectors of historical comment data in each application scenario;
a similarity parameter obtaining module 303, configured to calculate a similarity parameter for each application scene by using the feature vector;
an application scene preference degree generating module 304, configured to generate a preference degree of each application scene by using the user feature data and the similarity parameter;
and a preference recommending module 305, configured to recommend based on the preference degrees of the application scenarios.
Referring to fig. 4, it shows a block diagram of another structure of the user preference recommending apparatus based on the embodiment of fig. 3, which is as follows:
a user characteristic data extraction module 301, configured to extract user characteristic data in each application scenario from historical data;
preferably, the user feature data extraction module 301. The method specifically comprises the following steps:
the comment subject distribution feature vector extraction submodule 3011 is configured to extract a comment subject distribution feature vector of the user by performing LDA processing on history comment data of the user in each application scenario in the history data;
the user search word vector obtaining sub-module 3012 is configured to obtain a user search word vector by performing word2vec processing on historical search data of a user in the historical data;
a first user feature generation submodule 3013, configured to combine user basic information in historical data, the comment subject distribution feature vector of the user, and the user search word vector to generate user feature data of the user for each application scenario;
preferably, the user feature data extraction module 301 further includes:
the enhanced collaborative recommendation data acquisition sub-module is used for acquiring enhanced collaborative recommendation data of other users in each application scene through marked high-credibility user data in the historical data;
a second user characteristic generation submodule, configured to combine user basic information in the historical data with the enhanced collaborative recommendation data to generate user characteristic data of the user for each application scenario
Preferably, the enhanced collaborative recommendation data obtaining sub-module includes:
the high-reliability user data acquisition unit is used for acquiring marked high-reliability user data in each application scene in the historical data;
the application scene preference degree acquiring unit is used for acquiring the preference degree of the high-credibility user to each application scene through high-credibility user data;
the historical behavior log acquiring unit is used for extracting the historical behavior log of the high-credibility user from the high-credibility user data;
a behavior vector feature generation unit, configured to generate, by using an alternating least squares algorithm, behavior vector features of the high-confidence user in the application scenes according to the historical behavior log of the high-confidence user;
and the preference degree acquisition unit is used for acquiring the preference degrees of other users in the historical data to the application scenes according to the preference degrees and the behavior vector characteristics.
Preferably, the user feature data extraction module 301 further includes:
the time preference characteristic acquisition submodule is used for combining the historical behavior log of the user in the historical data with the time weight to acquire the time preference characteristic of the historical behavior of the user;
and the third user characteristic generation submodule is used for generating user characteristic data of the user aiming at each application scene by combining the user basic information in the historical data and the time preference characteristics of the user historical behaviors.
A comment feature vector acquisition module 302, configured to acquire feature vectors of historical comment data in each application scenario;
a similarity parameter obtaining module 303, configured to calculate a similarity parameter for each application scene by using the feature vector;
an application scene preference degree generating module 304, configured to generate a preference degree of each application scene by using the user feature data and the similarity parameter;
preferably, the application scene preference generating module 304 includes:
an application scene preference degree generating submodule 3041, configured to generate, by using a preset nonlinear classification model and a sorting algorithm, a preference degree of each application scene according to the user feature data and the similarity parameter.
A user active time obtaining module 305, configured to obtain user active time in each application scenario;
preferably, the user active time obtaining module 305 includes:
the user historical behavior log obtaining submodule is used for obtaining a user historical behavior log in the historical data;
and the active time acquisition submodule is used for acquiring the average active period of the user, the active time period of the user and the active time of the user in each application scene according to the historical behavior log of the user.
And a preference recommending module 306, configured to recommend based on the preference degrees in the application scenarios.
Preferably, the preference recommending module 306 includes:
the application scene determination submodule 3061 is configured to determine at least one application scene according to a search word currently input by the user in an average user activity period, a user activity time period, and an active time of the user in the application scene;
the recommendation displaying submodule 3062 is configured to display the at least one application scene to the user according to the preference degree of the application scene.
An embodiment of the present invention further provides an apparatus, including: a processor, a memory, and a computer program stored on the memory and executable on the processor, wherein the processor implements one or more of the user preference recommendation methods described above when executing the program.
Embodiments of the present invention also provide a readable storage medium, and when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the user preference recommendation method.
Preferably, the method further comprises the following steps:
and the active time recommending module is used for recommending based on the user active time and the corresponding preference degree under each application scene.
In summary, the user preference recommendation method provided by the embodiment of the present invention is configured to extract, by the user feature data extraction module, user feature data in each application scenario from historical data; secondly, a comment feature vector acquisition module is used for acquiring feature vectors of historical comment data in each application scene; calculating similarity parameters for each application scene by using the feature vectors through a similarity parameter acquisition module; then, a similarity parameter obtaining module is used for generating preference degrees of each application scene by adopting the user characteristic data and the similarity parameters; the user active time acquisition module is used for acquiring user active time in each application scene; and finally, the preference recommending module is used for recommending the preference of the user based on the preference degrees of the application scenes based on the active time of the user and the corresponding preference degrees of the application scenes. According to the proposal, the purpose of recommending the preference application scenes to the user aiming at the active time of the user is more refined through the activity of the user, the preference of the user to each application scene and the similarity parameter of each application scene is realized. It has the following advantages:
one is as follows: the merchant dimension preference is converted into the scene dimension preference in a mode of carrying out weighted aggregation on the user merchant preference, and the user scene preference is determined more effectively and accurately.
Secondly, the application of the LDA theme vector in the text preference mining of the user is combined with the query of the user, and the application that the user directly expresses the preferred text data is maximized.
And thirdly, expanding and filling the sparse behavior of the user to the merchant by ALS, and cooperatively expanding the scene preference user by the reinforced behavior.
The recommendation method comprises the step of recommending search terms, merchants, commodities and the like under related scenes to a user. By recommending to the user more accurately, the user may be triggered more click actions and thus yield more order conversions.
For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
The algorithms and displays presented herein are not inherently related to any particular computer, virtual machine, or other apparatus. Various general purpose systems may also be used with the teachings herein. The required structure for constructing such a system will be apparent from the description above. Moreover, the present invention is not directed to any particular programming language. It is appreciated that a variety of programming languages may be used to implement the teachings of the present invention as described herein, and any descriptions of specific languages are provided above to disclose the best mode of the invention.
In the description provided herein, numerous specific details are set forth. It is understood, however, that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be interpreted as reflecting an intention that: that the invention as claimed requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.
Those skilled in the art will appreciate that the modules in the device in an embodiment may be adaptively changed and disposed in one or more devices different from the embodiment. The modules or units or components of the embodiments may be combined into one module or unit or component, and furthermore they may be divided into a plurality of sub-modules or sub-units or sub-components. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or elements of any method or apparatus so disclosed, may be combined in any combination, except combinations where at least some of such features and/or processes or elements are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
The various component embodiments of the invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It will be appreciated by those skilled in the art that a microprocessor or Digital Signal Processor (DSP) may be used in practice to implement some or all of the functions of some or all of the components in a payment information processing apparatus according to embodiments of the present invention. The invention may also be embodied as an apparatus or device program (e.g., a computer program and computer program product data) for carrying out a part or all of the methods described herein. Such programs implementing the present invention may be stored on computer-readable media or may be in the form of one or more signals. Such a signal may be downloaded from an internet website or provided on a carrier signal or in any other form.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The usage of the words first, second and third, etcetera do not indicate any ordering. These words may be interpreted as names.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (9)
1. A recommendation method, characterized in that the method comprises:
extracting user characteristic data under each application scene from historical data;
collecting characteristic vectors of historical comment data in each application scene;
calculating similarity parameters aiming at each application scene by adopting the characteristic vectors;
generating preference degrees of each application scene by adopting the user characteristic data and the similarity parameters;
recommending based on the preference degrees of the application scenes;
acquiring user active time in each application scene; recommending based on the user active time and the corresponding preference degree in each application scene;
the step of generating the preference degrees of the application scenes by adopting the user characteristic data and the similarity parameters comprises the following steps:
generating the preference degrees of the application scenes by utilizing a preset nonlinear classification model and a sequencing algorithm according to the user characteristic data and the similarity parameters;
the step of obtaining the user active time in each application scene includes:
acquiring a user historical behavior log in the historical data;
and acquiring the average active period of the user, the active time period of the user and the active time of the user in each application scene according to the historical behavior log of the user.
2. The method according to claim 1, wherein the step of extracting the user feature data in each application scenario from the historical data comprises:
LDA processing is carried out on historical comment data of users in the historical data under various application scenes, and comment subject distribution characteristic vectors of the users are extracted;
obtaining a user search word vector by performing word2vec processing on historical search data of a user in the historical data;
and generating user characteristic data of the user aiming at each application scene by combining the user basic information in the historical data, the comment subject distribution characteristic vector of the user and the user search word vector.
3. The method according to claim 1, wherein the step of extracting the user feature data in each application scenario from the historical data comprises:
acquiring enhanced collaborative recommendation data of other users in each application scene through marked high-credibility user data in the historical data;
and combining the user basic information in the historical data with the strengthened collaborative recommendation data to generate user characteristic data of the user aiming at each application scene.
4. The method according to claim 1, wherein the step of extracting the user feature data in each application scenario from the historical data comprises:
combining the historical behavior log of the user in the historical data with the time weight to obtain the time preference characteristics of the historical behavior of the user;
and generating user characteristic data of the user aiming at each application scene by combining the user basic information in the historical data and the time preference characteristics of the user historical behaviors.
5. The method according to claim 3, wherein the step of obtaining the enhanced collaborative recommendation data for the user in each application scenario through the marked high-credibility user data in the historical data comprises:
acquiring marked high-reliability user data under each application scene in historical data;
acquiring the preference degree of the high-reliability user to each application scene through high-reliability user data;
extracting historical behavior logs of the high-credibility users from the high-credibility user data;
generating behavior vector characteristics of the high-reliability user in each application scene according to the historical behavior log of the high-reliability user by adopting an alternating least square algorithm;
and acquiring the preference degrees of other users in the historical data to each application scene according to the preference degrees and the behavior vector characteristics.
6. The method according to claim 1, wherein the step of recommending based on the preference of each application scenario comprises:
determining at least one application scene according to search words currently input by a user in an average user active period, a user active time period and active time of the user in the application scene;
and displaying the at least one application scene to the user according to the preference degree of the application scene.
7. A user preference recommendation apparatus, the apparatus comprising:
the user characteristic data extraction module is used for extracting user characteristic data under each application scene from historical data;
the comment feature vector acquisition module is used for acquiring feature vectors of historical comment data in each application scene;
a similarity parameter obtaining module for calculating similarity parameters for each application scene by using the feature vectors;
the application scene preference degree generating module is used for generating preference degrees of all application scenes by adopting the user characteristic data and the similarity parameters;
the preference recommending module is used for recommending based on the preference degrees of the application scenes;
the user active time acquisition module is used for acquiring user active time in each application scene;
the active time recommending module is used for recommending based on the user active time and the corresponding preference degree under each application scene;
the application scene preference generating module comprises:
an application scene preference degree generation submodule, configured to generate a preference degree of each application scene by using a preset nonlinear classification model and a sorting algorithm through the user feature data and the similarity parameter;
the user active time obtaining module comprises:
the user historical behavior log obtaining submodule is used for obtaining a user historical behavior log in the historical data;
and the active time acquisition submodule is used for acquiring the average active period of the user, the active time period of the user and the active time of the user in each application scene according to the historical behavior log of the user.
8. An electronic device, comprising:
processor, memory and computer program stored on the memory and executable on the processor, characterized in that the processor implements the recommendation method according to any of claims 1-6 when executing the program.
9. A readable storage medium, characterized in that instructions in the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the recommendation method according to any of method claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810843390.7A CN109299994B (en) | 2018-07-27 | 2018-07-27 | Recommendation method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810843390.7A CN109299994B (en) | 2018-07-27 | 2018-07-27 | Recommendation method, device, equipment and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109299994A CN109299994A (en) | 2019-02-01 |
| CN109299994B true CN109299994B (en) | 2021-09-24 |
Family
ID=65172668
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810843390.7A Active CN109299994B (en) | 2018-07-27 | 2018-07-27 | Recommendation method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109299994B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111800569B (en) * | 2019-04-09 | 2022-02-22 | Oppo广东移动通信有限公司 | Photographing processing method and device, storage medium and electronic equipment |
| CN110866180B (en) * | 2019-10-12 | 2022-07-29 | 平安国际智慧城市科技股份有限公司 | Resource recommendation method, server and storage medium |
| CN110825966B (en) * | 2019-10-31 | 2022-03-04 | 广州市百果园信息技术有限公司 | Information recommendation method and device, recommendation server and storage medium |
| CN111080339B (en) * | 2019-11-18 | 2024-01-30 | 口口相传(北京)网络技术有限公司 | Scene-based category preference data generation method and device |
| CN111159393B (en) * | 2019-12-30 | 2023-10-10 | 电子科技大学 | Text generation method for abstract extraction based on LDA and D2V |
| CN111368202B (en) * | 2020-03-06 | 2023-09-19 | 咪咕文化科技有限公司 | Search recommendation method and device, electronic equipment and storage medium |
| CN111553748B (en) * | 2020-05-09 | 2022-07-01 | 福州大学 | Android micro-service recommendation method and system based on user scene |
| CN112749343B (en) * | 2021-01-22 | 2023-04-07 | 武汉蔚来能源有限公司 | Resource recommendation method and device and computer storage medium |
| CN113496432B (en) * | 2021-07-06 | 2024-09-13 | 北京爱笔科技有限公司 | Mining method, device, equipment and storage medium for entity to be recommended |
| WO2023019517A1 (en) * | 2021-08-19 | 2023-02-23 | 阿波罗智联(北京)科技有限公司 | Instruction recommendation method and apparatus |
| CN114398486B (en) * | 2022-01-06 | 2022-08-26 | 北京博瑞彤芸科技股份有限公司 | Method and device for intelligently customizing customer acquisition publicity |
| CN114528484A (en) * | 2022-01-26 | 2022-05-24 | 北京金堤科技有限公司 | Preference mining method and device, storage medium and electronic equipment |
| CN114443967B (en) * | 2022-04-08 | 2022-07-08 | 北京并行科技股份有限公司 | Similar application recommendation method, computing device and storage medium |
| CN115470414B (en) * | 2022-11-03 | 2023-05-12 | 安徽商信政通信息技术股份有限公司 | Method and system for recommending joint persons |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103324750A (en) * | 2013-07-04 | 2013-09-25 | 莫志鹏 | Method for personal screening of photo galleries on the basis of Bayesian network |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103106285B (en) * | 2013-03-04 | 2017-02-08 | 中国信息安全测评中心 | Recommendation algorithm based on information security professional social network platform |
| CN105812937B (en) * | 2014-12-30 | 2019-05-24 | Tcl集团股份有限公司 | A kind of TV programme suggesting method and television program recommending device |
| US20180174218A1 (en) * | 2016-12-19 | 2018-06-21 | Sap Se | Recommendation optmization with a dynamic mixture of frequent and occasional recommendations |
| CN108230007B (en) * | 2017-11-28 | 2021-09-10 | 北京三快在线科技有限公司 | User intention identification method and device, electronic equipment and storage medium |
| CN108197285A (en) * | 2018-01-15 | 2018-06-22 | 腾讯科技（深圳）有限公司 | A kind of data recommendation method and device |
-
2018
- 2018-07-27 CN CN201810843390.7A patent/CN109299994B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103324750A (en) * | 2013-07-04 | 2013-09-25 | 莫志鹏 | Method for personal screening of photo galleries on the basis of Bayesian network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109299994A (en) | 2019-02-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109299994B (en) | Recommendation method, device, equipment and readable storage medium | |
| CN111784455B (en) | Article recommendation method and recommendation equipment | |
| CN108959603B (en) | Personalized recommendation system and method based on deep neural network | |
| CN107908740B (en) | Information output method and device | |
| KR101419504B1 (en) | System and method providing a suited shopping information by analyzing the propensity of an user | |
| CN103870973B (en) | Information push, searching method and the device of keyword extraction based on electronic information | |
| CN104899273B (en) | A kind of Web Personalization method based on topic and relative entropy | |
| CN109684538A (en) | A kind of recommended method and recommender system based on individual subscriber feature | |
| CN108182621A (en) | The Method of Commodity Recommendation and device for recommending the commodity, equipment and storage medium | |
| CN108805598B (en) | Similarity information determination method, server and computer-readable storage medium | |
| CN105718184A (en) | Data processing method and apparatus | |
| US9767417B1 (en) | Category predictions for user behavior | |
| CN107730346A (en) | The method and apparatus of article cluster | |
| CN110827112B (en) | Deep learning commodity recommendation method and device, computer equipment and storage medium | |
| CN111400613A (en) | Article recommendation method, device, medium and computer equipment | |
| US9767204B1 (en) | Category predictions identifying a search frequency | |
| CN111225009B (en) | Method and device for generating information | |
| CN111429161B (en) | Feature extraction method, feature extraction device, storage medium and electronic equipment | |
| CN104077417A (en) | Figure tag recommendation method and system in social network | |
| US10474670B1 (en) | Category predictions with browse node probabilities | |
| Lin et al. | A consumer review-driven recommender service for web e-commerce | |
| CN111104590A (en) | Information recommendation method, device, medium and electronic equipment | |
| Tayal et al. | Personalized ranking of products using aspect-based sentiment analysis and Plithogenic sets | |
| TW201719569A (en) | Identifying social business characteristic user | |
| CN111429214B (en) | Transaction data-based buyer and seller matching method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |