CN111078997B - Information recommendation method and device - Google Patents
Information recommendation method and device Download PDFInfo
- Publication number
- CN111078997B CN111078997B CN201911129660.9A CN201911129660A CN111078997B CN 111078997 B CN111078997 B CN 111078997B CN 201911129660 A CN201911129660 A CN 201911129660A CN 111078997 B CN111078997 B CN 111078997B
- Authority
- CN
- China
- Prior art keywords
- information
- user
- target
- users
- score
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
Abstract
The application provides an information recommendation method and device, relates to the technical field of Internet, and can improve accuracy and diversity of information recommendation results. The method comprises the following steps: acquiring a plurality of characteristic labels of a plurality of users in a user group, a plurality of attribute labels of a plurality of information in an information group, interaction information of the users in the user group and interaction information of the information in the information group; the interactive information of the user is used for representing a plurality of information browsed by the user, and the interactive information of the information is used for representing that the information is browsed by the plurality of users; constructing a user-information preference model by utilizing interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user; the user-information preference model is used for recommending information for a first target user; constructing an information-user audience model by utilizing the interactive information and the feature labels of the first target information in the information set; recommending the first target information to the users in the user group by using the information-user audience model.
Description
Technical Field
The application relates to the technical field of Internet, in particular to an information recommendation method and device.
Background
At present, most mobile phones APP realize a message pushing function, but when the commercial bank mobile phones APP provides information pushing service for clients, the same information is pushed to all clients, and the same information may not be content of interest to all clients. This results in some users needing to search for information themselves, and although the mobile phone APP of the existing bank increases the viscosity of the user by providing information search service, the financial information is of various kinds and contents, so that the user can hardly find the content of interest from the massive information. The problem of timeliness and accuracy of information searching by the user is questionable, which brings trouble to the user, reduces user experience and limits better development of the mobile phone APP of the bank.
In this regard, some banking cell phones APP now employ collaborative filtering algorithms to push information. I.e., users who have selected the same product are considered to have similar preferences, products selected by neighbor users who have selected the same product as the target user are recommended to the target user. However, the method for recommending information by adopting the collaborative filtering algorithm is not suitable for new users without history browsing records, and the data amount is large when neighbor user selection screening is performed, so that the realization is complicated. The other bank mobile phones APP conduct information recommendation by adopting an algorithm based on information content recommendation. I.e. it is believed that the user would like to have a similar product to that historically selected by the user, i.e. recommending to the target user a similar product to that historically selected by the target user. However, the algorithm for recommending based on the information content cannot cope with the use situation of the new user, and the recommendation based on the information content may result in the recommended information content being single and not meeting the requirements of the clients. Therefore, how to better recommend information by the mobile phone APP of the bank becomes a problem to be solved urgently.
Disclosure of Invention
The application provides an information recommendation method and device, which can combine user interest preference and information audience characteristics to recommend information to a target user so as to improve the accuracy and diversity of information recommendation content.
In order to achieve the above purpose, the application adopts the following technical scheme:
in a first aspect, the present application provides an information recommendation method, which may include:
acquiring a plurality of characteristic labels of a plurality of users in a user group, a plurality of attribute labels of a plurality of information in an information group, interaction information of the users in the user group and interaction information of the information in the information group; the interactive information of the user is used for representing a plurality of information browsed by the user, and the interactive information of the information is used for representing that the information is browsed by the plurality of users;
constructing a user-information preference model by utilizing interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user; the user-information preference model is used for recommending information for a first target user;
constructing an information-user audience model by utilizing the interactive information and the feature labels of the first target information in the information set; recommending the first target information to the users in the user group by using the information-user audience model.
In a second aspect, the present application provides an information recommendation apparatus, comprising: the system comprises an acquisition module, a construction module and a recommendation module. The acquisition module is used for acquiring a plurality of characteristic labels of a plurality of users in the user group, a plurality of attribute labels of a plurality of information in the information group, interaction information of the users in the user group and interaction information of the information in the information group; the interactive information of the user is used for representing a plurality of information browsed by the user, and the interactive information of the information is used for representing that the information is browsed by the plurality of users. The construction module is used for constructing a user-information preference model by utilizing the interaction information of the first target user in the user group and attribute tags of a plurality of information browsed by the first target user; an information-user audience model is constructed using the interactive information and the feature tags of the first target information in the information set. The recommendation module is used for recommending information for a first target user by using the user-information preference model and recommending the first target information to users in the user group by using the information-user audience model.
In a third aspect, the present application provides an information recommendation apparatus, comprising: a processor and a memory. Wherein the memory is used to store one or more programs. The one or more programs include computer-executable instructions that, when executed by the apparatus, cause the apparatus to perform the information recommendation method of the first aspect and any of its various alternative implementations.
In a fourth aspect, the present application provides a computer-readable storage medium having instructions stored therein, which when executed by a computer, perform the information recommendation method according to any of the above-described first aspect and its various alternative implementations.
In a fifth aspect, the present application provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform the information recommendation method according to any of the above-mentioned first aspect and its various alternative implementations.
In a sixth aspect, the present application provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform the information recommendation method according to any of the above-mentioned first aspect and its various alternative implementations.
The information recommendation method and device provided by the application are used for acquiring a plurality of characteristic labels of a plurality of users in a user group, a plurality of attribute labels of a plurality of information in an information group, interaction information of the users in the user group and interaction information of the information in the information group; the interactive information of the user is used for representing a plurality of information browsed by the user, and the interactive information of the information is used for representing that the information is browsed by the plurality of users; constructing a user-information preference model by utilizing interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user; the user-information preference model is used for recommending information for a first target user; constructing an information-user audience model by utilizing the interactive information and the feature labels of the first target information in the information set; recommending the first target information to the users in the user group by using the information-user audience model. Compared with the prior art, the information recommendation is single and the data operation amount is large. According to the information recommendation method provided by the application, firstly, a plurality of information of the user and the information, such as the feature marks of the user, the interaction information of the user, the attribute labels of the information and the interaction information of the information, are used, and because the information is generated in the process that the user actually browses the information, the information can better represent the preference of the user to the information or better represent a certain piece of information which is more suitable for the user, so that the accuracy of a recommendation result can be ensured. And secondly, the constructed user-information preference model and the constructed information-user audience model are utilized to directly recommend information to the user, and analysis and calculation are not required to be carried out every time when target neighbor user browsing histories with similar characteristics to the user are searched, so that the information recommendation efficiency can be improved. In addition, the information recommendation is not carried out solely by depending on the browsing history of the user, so that the diversity of recommendation results can be further ensured, and the problem of recommendation of new users and new information is solved.
Drawings
FIG. 1 is a diagram illustrating a first information recommendation method according to an embodiment of the present application;
FIG. 2 is a diagram illustrating a second information recommendation method according to an embodiment of the present application;
FIG. 3 is a diagram illustrating a third exemplary method for recommending information according to an embodiment of the present application;
FIG. 4 is a schematic diagram of an information recommendation apparatus according to an embodiment of the present application;
FIG. 5 is a schematic diagram of a second embodiment of an information recommendation apparatus;
fig. 6 is a schematic diagram of a structure of an information recommendation apparatus according to an embodiment of the present application.
Detailed Description
The information recommendation method and apparatus provided by the embodiments of the present application are described in detail below with reference to the accompanying drawings.
The terms "first" and "second" and the like in the description and in the drawings are used for distinguishing between different objects or between different processes of the same object and not for describing a particular order of objects.
Furthermore, references to the terms "comprising" and "having" and any variations thereof in the description of the present application are intended to cover a non-exclusive inclusion. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those listed but may optionally include other steps or elements not listed or inherent to such process, method, article, or apparatus.
It should be noted that, in the embodiments of the present application, words such as "exemplary" or "such as" are used to mean serving as an example, instance, or illustration. Any embodiment or design described herein as "exemplary" or "e.g." in an embodiment should not be taken as preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "such as" is intended to present related concepts in a concrete fashion.
In the description of the present application, unless otherwise indicated, the meaning of "a plurality" means two or more.
An embodiment of the present application provides an information recommendation method, as shown in fig. 1, the method may include S101-S104:
s101, the platform acquires a plurality of characteristic labels of a plurality of users in the user group, a plurality of attribute labels of a plurality of information in the information group, interaction information of the users in the user group and interaction information of the information in the information group.
Specifically, the platform may be a big data processing platform or other platforms with functions of data collection, processing, pushing, and the like, which is not particularly limited in the embodiment of the present application.
The user group consists of users using a certain APP, and the users include old users and new users, i.e. part of the users in the user group are old users with possible history browsing records, and part of the users are new users with possible no history browsing records. The feature labels of the users are the features of each user in the user group, and can be illustratively multidimensional information such as gender, age, academic, occupation, asset condition, liability condition, commercial product purchase condition and the like of the user. It will be appreciated that each user has at least one feature tag, and the scope of the feature tag is not particularly limited in the embodiments of the present application.
The information set is composed of the information in the APP information database, and the information includes old information and new information, namely, part of the information in the information set is the old information browsed by the user, and part of the information is the new information not browsed by the user. The attribute tag of information is an attribute feature to which each piece of information in the information set belongs, and each piece of information may contain one or more aspects of content, for example, one piece of information contains both political content and financial content. By way of example, the attribute tags may be multi-dimensional information such as politics, finance, entertainment, the internet, gardening, and the like. It will be appreciated that each piece of information has at least one attribute tag, and the scope of the attribute tag is not particularly limited in the embodiments of the present application.
The feature tag and the information tag may be manually marked or automatically extracted by using a natural language processing technology, which is not particularly limited in the embodiment of the present application.
The interactive information of the user is used for representing a plurality of information browsed by the user, namely, the information browsing condition of each user. When a user browses a piece of information, a history record is left, which indicates that the user browses the piece of information, and a feature tag of the user and an attribute tag of the information are also left. It will be appreciated that only old users who have browsed the information will have the user's interaction information.
The interactive information of the information is used to characterize the situation that the information is browsed by a plurality of users, i.e. each piece of information is browsed. It is also possible to know which users each piece of information has been browsed by, and the attribute tags of the piece of information and the feature tags of the users who have browsed the piece of information based on the history. It will be appreciated that only information viewed by the user has information interaction information.
S102, constructing a user-information preference model by utilizing interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user.
As a possible implementation manner, taking the first target user history browsing n pieces of information as shown in fig. 2, S102 may be implemented as the following steps S1021 to S1022:
s1021, constructing a user-information preference matrix by utilizing the interaction information of the first target user in the user group and attribute tags of a plurality of information browsed by the first target user.
The first target user is a user with interaction information of the user in the user group. The first target user u historically browses n pieces of information, and can determine that the n pieces of information correspond to m attribute tags based on the attribute tags obtained in step S101, then a user-information preference matrix X can be constructed u The method comprises the following steps:
wherein n represents that the first target user browses n pieces of information, m tablesShowing n pieces of information having m attribute tags in total, x ij Representing the first target user's score for the ith piece of information,
n、m、i、j、/>are all positive integers.
Alternatively to this, the method may comprise,the number of times the first target user u browses the ith piece of information can be expressed.
Illustratively, there are a total of 4 attribute tags, economic, recreational, military, political, for the 4 th attribute tag (e.g., political), if the 3 rd piece of information has the 4 th attribute tag, e.g., the content of the 3 rd piece of information includes a "politically" related keywordRepresenting the number of times user u browses the 3 rd piece of information, the first target user scoring the 3 rd piece of information is +.>The 2 nd piece of information does not have the 4 th attribute tag (for example, the 2 nd piece of information is entertainment type, the content of which does not contain politically related words), then +_>The value of (2) is 0, and the score of the first target user on the 3 rd piece of information is 0. The scoring mechanism mentioned above counts the number of times the user browses a certain information, i.e. when the information has a certain tag, the number of times the information browses under the tag can be regarded as the score of the information. In addition to the scoring mechanisms referred to above, it may also be entered manually by the user. In particular, the scoring of the information by the first target user may be generated by some scoring mechanism in the APP, So that after the user browses the information, the grading of the information can be input on the interface according to the preference degree of the user on the information. The score may be a ten-way score or a percentage-way score, which is not particularly limited in the embodiments of the present application.
Of course, the information may also be scored in combination with the various scoring mechanisms described above. For example, by combining the two scoring mechanisms, weights are set for two scores generated by the two scoring mechanisms respectively for a certain piece of information belonging to a certain label, and the two scores are weighted to obtain the final score of the information under the label.
Exemplary, the first target user u historically browses 4 pieces of information v 1 ，v 2 ，v 3 ，v 4 The 4 pieces of information are provided with 4 kinds of attribute tags s in total 1 ，s 2 ，s 3 ，s 4 Wherein the information v 1 With attribute tags s 1 Sum s 3 Information v 2 With attribute tags s 1 Sum s 2 Information v 3 With attribute tags s 3 Information v 4 With attribute tags s 2 Sum s 4 . The first target user u pair information v 1 The score of (2) is 8, the first target user u is about the information v 2 The score of (2) is 6, the first target user u is about the information v 3 The score of (2) is 10, the first target user u is relative to the information v 4 Is rated as 2 points. The score herein may be a score generated by any of the scoring mechanisms described above.
Thus, the user-information preference matrix for the first target user u is:
s1022, constructing a user-information preference model by using the user-information preference matrix.
Specifically, the process of constructing a user-information preference model using a user-information preference matrix is a solution to the multi-attribute decision problem. There are three main bodies in this multi-attribute decision problem: alternative set (thA set of all information that the target user u has historically browsed) v= { V 1 ,v 2 ,…,v i ,…,v n Attribute set of measurement scheme (set of attribute tags of all information) s= { S 1 ,s 2 ,…,s j ,…,s m Attribute weight vector representing attribute importance (preference of first target user to certain attribute tag)0≤w j ≤1，/>Alternative set v= { V 1 ,v 2 ,…,v i …,v n Each schema in may contain m dimensions (corresponding to m attribute tags). All alternatives of vectors can be combined to form a multi-attribute decision matrix, i.e. a user-information preference matrix X u . Wherein the attribute weight vector->In w j The weight of the j-th attribute tag in the m attribute tags representing the information. Illustratively, the information 1 includes attribute tags of politics, finance and the internet, and three attributes constitute all attributes of the information 1 content, wherein politics account for 70% of weight, finance accounts for 10% of weight and the internet accounts for 20% of weight in the three attribute tags.
As one possible implementation, in the multi-attribute decision theory, TOPSIS (technique for order preference by similarity to an ideal solution) method is a very effective multi-attribute decision method. The principle of TOPSIS is to order according to the proximity of a limited number of alternatives to an idealized target, which is an ordering approach that approximates to an ideal solution. In short, the basic principle of the TOPSIS method is to sequence the candidate solutions by detecting the distances between the candidate solutions and the optimal solution and between the candidate solutions and the worst solution, and if the candidate solutions are closest to the optimal solution and farthest from the worst solution, the candidate solutions are optimal. Wherein each index value of the optimal solution reaches the most of each evaluation indexAnd the optimal value and each index value of the worst solution reach the worst value of each evaluation index. Therefore, the TOPSIS method can be utilized to construct a user-information preference modelAnd uses the idea of TOPSIS approaching the ideal solution to find the optimal solution of the user-information preference model +.>Wherein the optimal solution of the user-information preference model can be expressed as +.>
Specifically, the construction of the user-information preference model by using the TOPSIS method can be realized by using the following steps:
step one, normalizing the user-information preference matrix to obtain a matrix E u ＝[e ij ] n×m 。
The normalization process has the advantages of improving the convergence rate of the model, improving the precision of the model and the like, so that the normalization process is firstly carried out on the user-information preference matrix so as to optimize the establishment of the subsequent model. The specific normalization processing method can be referred to the prior art, and the embodiment of the application is not specifically described.
Step two, matrix E u ＝[e ij ] n×m Weighting processing is carried out.
Since the importance of the attributes of different dimensions is different, the influence of the attribute weights needs to be considered in determining the ideal solution, and therefore, the normalized matrix E needs to be u ＝[e ij ] n×m Weighting is carried out, namely:
for the matrix Y in the TOPSIS method u Is typically a solution based on positive and negative ideal solutions:
for the embodiment of the application, the attribute labels of the information are neither benefit type attributes nor cost type attributes, namely, for the information, the attribute is not an element forming a certain optimal solution of the user-information preference model because the benefit generated by the corresponding attribute of the information is better, and the attribute is not an element forming a certain optimal solution of the user-information preference model because the corresponding cost of the attribute is lower. The solution described above based on positive and negative ideal solutions is not applicable to embodiments of the present application.
In the above step S1021, the score of the first target user for the information has been obtained, and as is known from the matrix (1), the scores of the users for the information having a certain same label also have a difference. Therefore, the user's score on the information can be utilized to obtain the preference degree of the user on the information attribute, namely, the average score corresponding to each attribute label can be obtained by utilizing the user's score on the information, namely, the preference degree of the user on a label. For example, the ideal solution can be approximatedExpressed as: />Wherein (1)>Ideal solution representing the required approximation, +.>Representing the optimal solution of the user-information preference model, < >>Attribute tags s representing information of a first target user u in a normalized user-information preference matrix j Average of multiple scores of (a). For example, in the above matrix (1), +.>I.e. for the current 4 pieces of information, the attribute tag s 2 The average score for user u was 2 points.
Step three, solving the optimal solution
The optimization model is built based on the TOPSIS method, i.e. the sum of squares of the distances between the optimal solution and the ideal solution needs to be minimized, then,
for convenience, letAnd a lagrangian Lagrange equation was constructed as follows:
Respectively toAnd λ and let equation be 0, namely:
solving equation (1) above yields the solution as follows:
for Lagrange equationIs->The second order partial derivative is calculated to construct a black plug Hesse matrix as follows:
due toTherefore->Is a positive definite matrix. Thus (S)>Is the optimal solution for user-information preferences.
Step four, constructing a user-information preference model
Solving the optimal solution according to the third stepThe method of (1) obtains the optimal solution corresponding to each attribute tag in m attribute tags, and can be expressed as +.>Further construct a user-information preference modelEach element in the user-information preference model can be best de-tabulatedShowing, i.e
S103, recommending information for the first target user by using the user-information preference model.
As one possible implementation, information whose preference degree satisfies the first condition is recommended to the first target user.
Specifically, a user-information preference model can be created for each user according to the method, and the description can be solved according to the ideal solutionThe historical score average of the user for certain information attribute tags and the user-information preference model of the user can be utilized to determine the score of the user for the information, wherein the score is the preference degree of the user for the information. At this time, the first condition may be set as information for which the score of the information by the user is higher than the first preset threshold, and then the information satisfying the first condition may be recommended to the user.
Illustratively, the user-information preference matrix for the first target user is:
the user-information preference model is: />At this time, the first information and the attribute tag included in the first information are acquired, assuming that the first information includes the attribute tag: s is(s) 1 Sum s 3 . First, an average score of attribute tags contained in each first information by a first target user is obtained. Then the first target user pair attribute tags s can be obtained by referring to the user-information preference matrix of the first target user 1 The average score of (8+6)/2=7, the first target user pairs the attribute tags s 3 The average score of (8+10)/2=9. Then, obtaining a first objective according to the average score and the user-information preference modelThe first score of the target user for the first information is +.>The final result of multiplying the historical average score of the attribute labels contained in the first information by the corresponding weight and then summing the multiplied historical average score is the first score. Assuming that the first preset threshold is 5 at this time and the first score is 5.1 > 5, the first information is recommended to the first target user. The first preset threshold may be obtained according to an empirical value, which is not limited in the embodiment of the present application. The recommended information is the information of the most preferred aspect of the user, so that the interest of the user can be increased, and the viscosity of the client can be increased. Even a new piece of information can be recommended to the corresponding user according to the attribute label of the information, so as to solve the problem of cold start of the new product.
Optionally, the first condition may be a preset number of information whose preference degree satisfies the preset condition, that is, the first scores higher than the first preset threshold may be ranked, and a certain number of information with a previous ranking may be recommended to the corresponding user, so that the situation that the satisfaction degree of the user is reduced due to that too many information is recommended to the user is avoided. Correspondingly, when the preference degree meets the preset condition (the first score is larger than the first preset threshold value), the information quantity is smaller than the preset quantity, and the information of all quantities is recommended to the corresponding users so as to meet the user requirements.
Alternatively, because the information has timeliness, when the first scores of some information are the same or similar, a certain amount of information with the latest release time in the information can be recommended to the user, so that higher-quality information can be recommended to the user, and the user satisfaction is improved.
S104, constructing an information-user audience model by using the interactive information and the feature labels of the first target information in the information set.
As a possible implementation manner, taking the example that d users browse in the first target information history, as shown in fig. 3, S104 may be implemented as the following steps S1041 to S1042:
S1041, constructing an information-user audience matrix by utilizing the interactive information of the first target information in the information set and the feature labels of a plurality of users who browse the first target information.
Wherein the first target information history is browsed by d users, and c feature labels corresponding to the d users can be determined according to the feature labels obtained in step S101, so as to construct an information-user audience matrix Z v The method comprises the following steps:
wherein d represents that d users in the information group have browsed the first target information, c represents that d users have c feature tags in total, z ab Representing the score of the first target information by the a-th user,d、c、a、b、/>are all positive integers;
alternatively to this, the method may comprise,a score of the first target information by the user who browses the first target information may be represented. The score may be manually entered by the user, e.g., the score may be generated by some scoring mechanism in the APP, so that after the user has browsed the information, the user may score the piece of information according to his/her own preference for the piece of information. The score may be a ten-way score or a percentage-way score, which is not particularly limited in the embodiments of the present application. Of course, other scoring mechanisms may be employed to score information. Or combining a plurality of scoring mechanisms, and setting corresponding weights for each scoring mechanism to obtain the final score of the first target information under each label.
Illustratively, there are 4 users u in the history 1 ，u 2 ，u 3 ，u 4 Browsing the first target information, the 4 users having a total of 4 feature tags p 1 ，p 2 ，p 3 ，p 4 Wherein user u 1 With characteristic tag p 1 And p 3 User u 2 With characteristic tag p 1 And p 2 User u 3 With characteristic tag p 3 User u 4 With characteristic tag p 2 And p 4 . User u 1 The first target information is scored as 8 points, user u 2 The first target information is scored as 6 points, user u 3 The first target information is scored as 10 points, user u 4 The score of the first target information is 2 points. The score herein may be a score generated by any of the scoring mechanisms described above.
Thus, the information-user audience matrix for the first target information v is:
s1042, constructing an information-user audience model by using the information-user audience matrix.
Specifically, the above-mentioned information-user audience matrix establishment process is a multi-attribute decision-making process. There are three main bodies in this multi-attribute decision problem: alternative set (set of all users who have historically browsed the first target information v) u= { U 1 ,u 2 ,…,u a ,…,u d Attribute set of the measurement scheme (set of feature labels of all users) p= { P 1 ,p 2 ,…,p b ,…,p c Attribute weight vector representing attribute importance degree (degree of matching of first target information with a certain feature label) 0≤w b ≤1，/>Alternative set u= { U 1 ,u 2 ,…,u a ,…,u d Each of the schemes may contain c dimensions (corresponding to c feature labels). All the scheme vectors can be combined to form a multi-attribute decision matrix, namely an information-user audience matrix Z v . Wherein the attribute weight vectorIn w b The weight of the b-th feature tag in the c feature tags of the information is represented. Illustratively, the user 1 includes feature tags of 30% weight for men, 40% weight for programmers, and 50K for average month revenue, which constitute all feature attributes of the user 1.
As a possible implementation, the TOPSIS method described above can also be used to construct information-user audience modelsAnd using the idea of TOPSIS approach to approach the ideal solution to find the optimal solution of the information-user audience model>Wherein the optimal solution of the information-user audience model can be expressed as
Similar to the user-information preference model construction step, the construction of the information-user audience model using the TOPSIS method can be accomplished using the following steps:
step one, normalizing the information-user audience matrix to obtain a matrix E v ＝[e ab ] d×c 。
Step two, matrix E v ＝[e ab ] d×c Weighting processing is carried out.
I.e.
In the above step S1041, the score of the user for the first target information has been obtained, and it is known from the matrix (2) that the score of the user for the first target information with some identical feature tag also has a difference. Therefore, the degree of matching between the first target information and the user attribute tag can be obtained by using the score of the user on the first target information, namely, the average score corresponding to each feature tag can be obtained by using the score of the user on the first target information, namely, the degree of matching between the feature tag and the first target information. For example, the ideal solution can be approximatedExpressed as:wherein (1)>Ideal solution representing the required approximation, +.>Representing the optimal solution of the above information-user audience model,/->Representing normalized information-corresponding feature labels p calculated after scoring a first target information v by a user in a user audience matrix b Average of multiple scores of (a). For example, in the above matrix (2),i.e. for the current 4 users who have browsed the first target information v, the feature tag p 2 The average score of (2) was 2 points.
Step three, solving the optimal solution
See step S1022, above, for user-information preferences In the process of model establishment, the optimal solutionBased on the same algorithm idea, to obtain:
wherein, the liquid crystal display device comprises a liquid crystal display device,
fourth, build information-user audience model
Solving the optimal solution according to the third stepThe method of (2) obtaining the optimal solution corresponding to each of the c feature tags may be expressed as +.>Further construct a user-information preference model +.>Each element in the user-information preference model can be represented by an optimal solution, namely +.>
S105, recommending the first target information to the users in the user group by using the information-user audience model.
As one possible implementation, the first targeting information is recommended to users in the information-user audience model whose degree of matching satisfies the second condition.
Specifically, an information-user audience model may be created for each information according to the above method,at this time, from the above description of the ideal solution, it is known thatThe historical score averages for information for certain user feature tags may be utilized. That is, the average score of the information for the user with some feature tag in the history, for example, the historical score of the information for the user 1 aged 40 is 8, the historical score of the information for the user 2 aged 40 is 6, and the historical average score of the information for the feature tag aged 40 is (8+6)/2=7. And the information-user audience model of the information, determining the score of the information for the users, wherein the score is the matching degree of the information and the users. At this time, the second condition may be set to be that the score of the information to the users is higher than the second preset threshold, and then the information may be recommended to the users satisfying the second condition.
Illustratively, the information-user audience matrix for the first targeting information is:
the information-user audience model is:at this time, the first user and the feature tag contained in the first user are acquired, and it is assumed that the first user contains the feature tag: p is p 1 And p 3 . First, an average score of the first target information on the feature labels contained by each first user is obtained. Then the first target information pair feature tag p can be obtained by referring to the information-user audience matrix of the first target information 1 The average score of (8+6)/2=7, the first target information is against the feature tag p 3 The average score of (8+10)/2=9. Then, obtaining a second score of the first target information to the first user according to the average score and the information-user audience modelI.e. using the first target information to identify the feature tag contained in the first userThe final result of the summation of the historical average score multiplied by the corresponding weight is the second score. Assuming that the second preset threshold is 5 at this time and the second score is 5.1 > 5, the first target information is recommended to the first user. The second threshold may be obtained according to an empirical value, which is not limited in the embodiment of the present application. At this time, the user recommended by the information is the user with higher matching degree with the information, so that the interest of the user can be increased, and the viscosity of the client can be increased. Even a new user can recommend corresponding information to the user according to the characteristic label of the user, and the recommended information can meet the requirement of the client so as to solve the cold start problem of the new user.
Optionally, in order to avoid recommending excessive information to the same user, the second scores corresponding to a certain user above a second preset threshold may be ranked, and a certain amount of information with a previous ranking may be recommended to the user, so that the situation that the satisfaction of the user is reduced due to recommending excessive information to the user is avoided. Correspondingly, when the matching degree meets the preset condition (the second score is larger than the second preset threshold value), the information quantity is smaller than the preset quantity, and the information of all quantities is recommended to the corresponding users so as to meet the user requirements.
Optionally, because the information has timeliness, when the second scores of some information to the same user are the same or similar, a certain amount of information with the latest release time in the information can be recommended to the user, so that better information can be recommended to the user, and the user satisfaction is improved.
Further, through the above steps, the user-information preference model and the information-user audience model are established as the user recommended information, but in order to achieve a better recommending effect, the information recommending method according to the embodiment of the present application may further include the following step S106:
S106, according to the recommendation results of the user-information preference model and the information-user audience model, recommending the optimal E pieces of information to the first target user or recommending the first target information to the optimal F users.
Optionally, a first preset condition may be set according to an empirical value, for example, the user-information preference model accounts for 60% of the recommended result, and the information-user audience model accounts for 40% of the recommended result, so as to perform information fusion recommendation. Illustratively, assume that the first preset threshold and the second preset threshold are both 5; the first score of user A for information 1 is 8, and the second score of information 1 for user A is 6; the first score for user a for information 2 is 6 and the second score for information 2 for user a is 8. Then, if the fusion recommendation is not performed, both the information 1 and the information 2 can be recommended to the user A. If fusion recommendation is performed, the fusion score of information 1 is (8 x 60% +6 x 40%) =7.2; the fusion score for information 2 was (6 x 60% +8 x 40%) =6.8. Assuming that the preset threshold of the fusion recommendation score is 7, if the fusion score of the information 1 is greater than 7, recommending the information 1 to the user A, and if the fusion score of the information 2 is less than 7, not recommending the information 2 to the user A. By fusing the recommendation results of the user-information preference model and the information-user audience model, the accuracy and diversity of the recommendation results are further ensured. The information recommendation result ensures that the recommended information meets the preference degree of the user and meets the matching degree of the recommended information and the user.
Optionally, the number may be preset in the first preset condition, scores greater than a preset threshold in the fusion scores of all the information about a certain user are ranked, a certain number of information with a previous ranking is selected and recommended to the user, or scores greater than a preset threshold in the fusion scores of all the users about a certain information are ranked, a certain number of users with a previous ranking is selected and recommended to the users. That is, E pieces of information are preferentially selected from the recommendation results of the two models, and are recommended to the first target user, or F users are preferentially selected, and the first target information is recommended to the F users. Wherein E and F may be positive integers set according to empirical values. This avoids the situation that too much information is recommended to the user to reduce the satisfaction of the user. Correspondingly, when the information quantity of which the fusion score meets the preset condition (larger than a preset threshold value) is smaller than the preset quantity, recommending all the information to the corresponding user so as to meet the user requirement.
Optionally, because the information has timeliness, when the fusion scores of some information are the same or similar, a certain amount of information with the latest release time in the information can be recommended to the corresponding user, so that better information can be recommended to the user, and the user satisfaction is improved.
The information recommendation method provided by the application comprises the steps of obtaining a plurality of characteristic labels of a plurality of users in a user group, a plurality of attribute labels of a plurality of information in an information group, interaction information of the users in the user group and interaction information of the information in the information group; the interactive information of the user is used for representing a plurality of information browsed by the user, and the interactive information of the information is used for representing that the information is browsed by the plurality of users; constructing a user-information preference model by utilizing interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user; the user-information preference model is used for recommending information for a first target user; constructing an information-user audience model by utilizing the interactive information and the feature labels of the first target information in the information set; recommending the first target information to the users in the user group by using the information-user audience model. Compared with the prior art, the information recommendation is single and the data operation amount is large. The information recommending method provided by the application utilizes the constructed user-information preference model and the information-user audience model to recommend information for the user, can improve the efficiency, ensures the accuracy and the diversity of recommending results, and solves the recommending problem of new users and new information.
The embodiment of the application can divide the functional modules or functional units of the device according to the method example, for example, each functional module or functional unit can be divided corresponding to each function, or two or more functions can be integrated in one processing module. The integrated modules may be implemented in hardware, or in software functional modules or functional units. The division of the modules or units in the embodiment of the present application is schematic, which is merely a logic function division, and other division manners may be implemented in practice.
Fig. 4 shows a schematic diagram of one possible construction of the device involved in the above embodiment. The apparatus 400 includes an acquisition module 401, a construction module 402, and a recommendation module 403.
An obtaining module 401, configured to obtain a plurality of feature labels of a plurality of users in a user group, a plurality of attribute labels of a plurality of information in an information group, interaction information of users in the user group, and interaction information of information in the information group; the interactive information of the user is used for representing a plurality of information browsed by the user, and the interactive information of the information is used for representing that the information is browsed by the plurality of users.
A building module 402, configured to build a user-information preference model by using interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user; and the information-user audience model is constructed by utilizing the interactive information of the first target information in the information group and the characteristic label.
A recommendation module 403 for recommending information to a first target user using the user-information preference model and recommending the first target information to users in the user group using the information-user audience model.
In one possible implementation manner, the building module 402 is configured to build a user-information preference model by using the interaction information of the first target user in the user group and attribute tags of a plurality of information browsed by the first target user, which may specifically be:
constructing a user-information preference matrix by utilizing interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user; the user-information preference matrix X u The method comprises the following steps:
wherein n represents that the first target user browses n pieces of information, m represents that the n pieces of information have m attribute tags, x ij Representing the first target user's score for the ith piece of information, n、m、i、j、Are all positive integers;
constructing a user-information preference model using a user-information preference matrix0≤w j ≤1，/>Wherein w is j Indicating the preference degree of the first target user for the j-th attribute tag.
The building module 402 is configured to build an information-user audience model by using the interactive information and the feature tag of the first target information in the information set, which may specifically be:
constructing an information-user audience matrix by utilizing the interactive information of the first target information in the information set and the characteristic labels of a plurality of users who browse the first target information; the information-user audience matrix Z v The method comprises the following steps:
wherein d represents that d users in the information group have browsed the first target information, c represents that d users have c feature tags, z ab Representing the score of the first target information by the a-th user, d、c、a、b、/>are all positive integers;
utilization information-user receptionInformation-user audience model constructed by audience matrix0≤w b ≤1，/>Wherein (1)>Indicating the matching degree of the first target information and the b-th feature tag.
In one possible implementation, the recommendation module 403 is configured to recommend information to the first target user using the user-information preference model and recommend the first target information to the user in the user group using the information-user audience model. The method specifically comprises the following steps:
Recommending information with preference degree meeting a first condition for a first target user by using the user-information preference model;
recommending the first target information to the users in the user group by using the information-user audience model comprises the following steps: and recommending the first target information to the user with the matching degree meeting the second condition.
In one possible implementation, the recommending module 403 uses the user-information preference model to recommend information whose preference degree meets the first condition for the first target user, including: the first information and the attribute tag contained in the first information are obtained. An average score of the attribute tags contained in each first information by the first target user is obtained. A first score for the first information is obtained for the first target user based on the average score and the user-information preference model. If the first score is higher than a first preset threshold, the first information is information with the preference degree meeting the first condition, and the first information is recommended to the first target user. Recommending the first target information to the user with the matching degree meeting the second condition by using the information-user audience model, wherein the method comprises the following steps: the method comprises the steps of obtaining a first user and a feature tag contained by the first user. An average score of the first target information for the feature tags included by each first user is obtained. A second score for the first user is obtained for the first target information based on the average score and the information-user audience model. If the second score is higher than a second preset threshold, the first user recommends the first target information to the first user as the user with the matching degree meeting the second condition.
In a possible implementation manner, the recommendation module 403 is further configured to combine the recommendation results of the user-information preference model and the information-user audience model according to a first preset condition, and recommend the optimal E pieces of information to the first target user or recommend the first target information to the optimal F users; wherein E and F are positive integers.
The information recommending device provided by the application acquires a plurality of characteristic labels of a plurality of users in a user group, a plurality of attribute labels of a plurality of information in an information group, interactive information of the users in the user group and interactive information of the information in the information group; the interactive information of the user is used for representing a plurality of information browsed by the user, and the interactive information of the information is used for representing that the information is browsed by the plurality of users; constructing a user-information preference model by utilizing interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user; the user-information preference model is used for recommending information for a first target user; constructing an information-user audience model by utilizing the interactive information and the feature labels of the first target information in the information set; recommending the first target information to the users in the user group by using the information-user audience model. Compared with the prior art, the information recommendation is single and the data operation amount is large. The information recommending device provided by the application utilizes the constructed user-information preference model and the information-user audience model to recommend information for the user, can improve the efficiency, ensures the accuracy and the diversity of the recommending result, and solves the recommending problem of new users and new information.
FIG. 5 is a schematic diagram showing a possible structure of the information recommending apparatus. As shown in fig. 5, the apparatus 500 may include: a processor 501 and a communication interface 502. The processor 501 is configured to control and manage the actions of the apparatus 500, for example, performing the steps performed by the acquisition module 401, the construction module 402, and the recommendation module 403 described above, and/or to perform other processes of the techniques described herein. The communication interface 502 is used to support communication of the device with other network entities. The apparatus 500 may further comprise a memory 503 and a bus 504, the memory 503 being for storing program codes and data for the apparatus.
Wherein the memory 503 may comprise volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk or solid state disk; the memory may also comprise a combination of the above types of memories.
The processor 501 may be implemented or executed with the various exemplary logic blocks, modules and circuits described in connection with the present disclosure. The processor may be a central processing unit, a general purpose processor, a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic device, a transistor logic device, a hardware component, or any combination thereof. Which may implement or perform the various exemplary logic blocks, modules and circuits described in connection with this disclosure. The processor may also be a combination that performs the function of a computation, e.g., a combination comprising one or more microprocessors, a combination of a DSP and a microprocessor, etc.
Bus 504 may be an extended industry standard architecture (extended industry standard architecture, EISA) bus or the like. The bus 504 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in fig. 5, but not only one bus or one type of bus.
As shown in FIG. 6, another possible structure of the information recommendation apparatus is provided in the embodiment of the present application. The apparatus 600 includes: a processing unit 601. The processing unit 601 is configured to control and manage the actions of the apparatus 400, for example, perform the steps performed by the acquisition module 401, the construction module 402, the recommendation module 403, and/or perform other processes of the techniques described herein. The apparatus may further comprise a storage unit 602 and a communication unit 603, the storage unit 602 being configured to store program codes and data of the apparatus; the communication unit 603 is configured to support communication of the device with other network entities.
Wherein, in connection with fig. 5 and 6, the processing unit 601 may be the processor 501 or the controller in the apparatus 500.
The memory unit 602 may be a memory or the like in the apparatus 500, which may include a volatile memory, such as a random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk or solid state disk; the memory may also comprise a combination of the above types of memories.
The communication unit 603 may be a communication interface 502 in the apparatus 500, or a transceiver, transceiving circuitry, etc.
From the foregoing description of the embodiments, it will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of functional modules is illustrated, and in practical application, the above-described functional allocation may be implemented by different functional modules according to needs, i.e. the internal structure of the apparatus is divided into different functional modules to implement all or part of the functions described above. The specific working processes of the above-described systems, devices and units may refer to the corresponding processes in the foregoing method embodiments, which are not described herein.
The embodiment of the application also provides a computer readable storage medium, wherein the computer readable storage medium stores instructions, and when the information recommending device executes the instructions, the device executes each step executed by the information recommending device in the method flow shown in the method embodiment.
The computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (random access memory, RAM), a read-only memory (ROM), an erasable programmable read-only memory (erasable programmable read only memory, EPROM), a register, a hard disk, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing, or any other form of computer readable storage medium known in the art. An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an application specific integrated circuit (application specific integrated circuit, ASIC). In embodiments of the present application, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The foregoing is merely illustrative of specific embodiments of the present application, and the scope of the present application is not limited thereto, but any changes or substitutions within the technical scope of the present application should be covered by the scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
Claims (8)
1. An information recommendation method, comprising:
acquiring a plurality of characteristic labels of a plurality of users in a user group, a plurality of attribute labels of a plurality of information in an information group, interaction information of the users in the user group and interaction information of the information in the information group; the interactive information of the user is used for representing a plurality of information browsed by the user, and the interactive information of the information is used for representing that the information is browsed by a plurality of users;
constructing a user-information preference model by utilizing interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user; the user-information preference model is used for recommending information for the first target user;
constructing an information-user audience model by utilizing the interactive information and the characteristic labels of the first target information in the information group; recommending the first target information to users in the user group by using the information-user audience model;
Constructing a user-information preference model by using the interaction information of the first target user in the user group and attribute tags of a plurality of information browsed by the first target user, wherein the method specifically comprises the following steps:
constructing a user-information preference matrix by utilizing interaction information of a first target user in the user group and attribute tags of a plurality of information browsed by the first target user; the user-information preference matrix X u The method comprises the following steps:
wherein n represents that the first target user browses n pieces of information, m represents that the n pieces of information have m attribute tags, x ij Representing a score of the first target user for the ith piece of information,n、m、i、j、/>are all positive integers;
constructing the user-information preference model using the user-information preference matrix,wherein w is j And representing the preference degree of the first target user on the j-th attribute label.
2. The information recommendation method according to claim 1, wherein the constructing an information-user audience model using the interactive information and the feature tags of the first target information in the information group comprises:
using the first target information in the information setConstructing an information-user audience matrix by the interactive information of a plurality of users and the characteristic labels of the users browsed by the first target information; the information-user audience matrix Z v The method comprises the following steps:
wherein d represents that d users in the information group have browsed the first target information, c represents that the d users have c feature tags, z ab Representing a score of the first target information by the a-th user,d、c、a、b、/>are all positive integers;
constructing the information-user audience model using the information-user audience matrixWherein w is b Indicating the matching degree of the first target information and the b-th feature tag.
3. The information recommendation method according to claim 1 or 2, wherein,
the user-information preference model is specifically used for recommending information with preference degree meeting a first condition for the first target user;
the information-user audience model is specifically used for recommending the first target information to the user with the matching degree meeting the second condition.
4. The information recommendation method according to claim 3, wherein said recommending information having a preference degree satisfying a first condition for said first target user comprises:
acquiring first information and attribute tags contained in the first information;
obtaining the average score of the attribute tags contained in each piece of first information by the first target user;
obtaining a first score of the first information by the first target user according to the average score and the user-information preference model;
If the first score is higher than a first preset threshold, the first information is the information of which the preference degree meets a first condition, and the first information is recommended to the first target user.
5. The information recommendation method according to claim 3, wherein said recommending the first target information to the user whose matching degree satisfies the second condition comprises:
acquiring a first user and a feature tag contained by the first user;
obtaining the average score of the first target information on the feature labels contained by each first user;
obtaining a second score of the first target information to the first user according to the average score and the information-user audience model;
if the second score is higher than a second preset threshold, the first user recommends the first target information to the first user as the user of which the matching degree meets a second condition.
6. The information recommendation method according to claim 4 or 5, wherein said method further comprises:
according to a first preset condition, fusing recommendation results of the user-information preference model and the information-user audience model, recommending optimal E pieces of information to the first target user or recommending the first target information to optimal F users; wherein E and F are positive integers.
7. An information recommendation device, characterized in that the information recommendation device comprises: a processor and a memory, wherein the memory is configured to store one or more programs, the one or more programs comprising computer-executable instructions, which when executed by the information recommendation device, cause the information recommendation device to perform the information recommendation method of any one of claims 1 to 6.
8. A computer-readable storage medium, wherein instructions are stored in the computer-readable storage medium, which when executed by a computer, perform the information recommendation method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911129660.9A CN111078997B (en) | 2019-11-18 | 2019-11-18 | Information recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911129660.9A CN111078997B (en) | 2019-11-18 | 2019-11-18 | Information recommendation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111078997A CN111078997A (en) | 2020-04-28 |
| CN111078997B true CN111078997B (en) | 2023-08-22 |
Family
ID=70311125
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911129660.9A Active CN111078997B (en) | 2019-11-18 | 2019-11-18 | Information recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111078997B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111538905B (en) * | 2020-04-29 | 2023-10-20 | 中国银行股份有限公司 | Object recommendation method and device |
| CN113836347A (en) * | 2020-06-23 | 2021-12-24 | 北京达佳互联信息技术有限公司 | Recommendation method and device and electronic equipment |
| CN113592198B (en) * | 2021-08-24 | 2023-09-26 | 腾讯科技（深圳）有限公司 | Method, server and terminal for determining demand reference information |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108711075A (en) * | 2018-05-22 | 2018-10-26 | 阿里巴巴集团控股有限公司 | A kind of Products Show method and apparatus |
| CN109543111A (en) * | 2018-11-28 | 2019-03-29 | 广州虎牙信息科技有限公司 | Recommendation information screening technique, device, storage medium and server |
| CN110457581A (en) * | 2019-08-02 | 2019-11-15 | 达而观信息科技（上海）有限公司 | A kind of information recommended method, device, electronic equipment and storage medium |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10699321B2 (en) * | 2017-10-17 | 2020-06-30 | Adobe Inc. | Global vector recommendations based on implicit interaction and profile data |
-
2019
- 2019-11-18 CN CN201911129660.9A patent/CN111078997B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108711075A (en) * | 2018-05-22 | 2018-10-26 | 阿里巴巴集团控股有限公司 | A kind of Products Show method and apparatus |
| CN109543111A (en) * | 2018-11-28 | 2019-03-29 | 广州虎牙信息科技有限公司 | Recommendation information screening technique, device, storage medium and server |
| CN110457581A (en) * | 2019-08-02 | 2019-11-15 | 达而观信息科技（上海）有限公司 | A kind of information recommended method, device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111078997A (en) | 2020-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10963541B2 (en) | Systems, methods, and apparatuses for implementing a related command with a predictive query interface | |
| US20220005125A1 (en) | Systems and methods for collecting and processing alternative data sources for risk analysis and insurance | |
| CN109033101B (en) | Label recommendation method and device | |
| CN109582876B (en) | Tourist industry user portrait construction method and device and computer equipment | |
| CN111078997B (en) | Information recommendation method and device | |
| CN106844407B (en) | Tag network generation method and system based on data set correlation | |
| WO2018040069A1 (en) | Information recommendation system and method | |
| CN103309869A (en) | Method and system for recommending display keyword of data object | |
| CN114371946B (en) | Information push method and information push server based on cloud computing and big data | |
| CN114819967A (en) | Data processing method and device, electronic equipment and computer readable storage medium | |
| WO2019144035A1 (en) | Systems and methods for collecting and processing alternative data sources for risk analysis and insurance | |
| CN110598126B (en) | Cross-social network user identity recognition method based on behavior habits | |
| US20150206220A1 (en) | Recommendation Strategy Portfolios | |
| CN109727056B (en) | Financial institution recommendation method, device, storage medium and device | |
| CN112685635A (en) | Item recommendation method, device, server and storage medium based on classification label | |
| US20230177581A1 (en) | Product metadata suggestion using embeddings | |
| You et al. | A hotel ranking model through online reviews with aspect-based sentiment analysis | |
| CN115033757A (en) | Service sequencing method and device, storage medium and electronic equipment | |
| CN117333191A (en) | Complaint event association method and device and complaint event association system | |
| CN116578767A (en) | Semantic data processing and content recommending method and device and computer equipment | |
| Kaul et al. | Evaluating Techniques for Mining Customer Purchase Behavior and Product Recommendation-A Survey | |
| CN113269610A (en) | Bank product recommendation method and device and storage medium | |
| CN117541350A (en) | Product pushing method, device, computer equipment and storage medium | |
| CN117131258A (en) | Audience data processing method and system based on data center | |
| CN114078057A (en) | Method, device and storage medium for determining enterprise financing probability |
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 |