Recommendation method and system based on AHP weight and trust degree model

Abstract

The application discloses a recommendation method and system based on AHP weight and trust degree model, and belongs to the technical field of collaborative filtering. The application carries out AHP weighting on the score difference of a training project to analyze weak subjects which have greater influence on users, and effectively improves the pertinence of the recommended training project. The similarity of users is calculated through a trust degree model to obtain similar users of the users, and finally, collaborative filtering priority recommendation is carried out. The provided recommendation mode can effectively alleviate the data sparsity problem in the process of collaborative filtering priority recommendation, and effectively improves the efficiency and accuracy of the collaborative filtering recommendation.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of collaborative filtering, and particularly relates to a recommendation method and system based on an AHP weight and a trust degree model. BACKGROUND

[0002] Since the Internet has entered the era of big data, the behavior of enterprises and users has undergone a series of changes and reconstruction. The biggest change is in the business, all user behaviors are "visualized". With the continuous development and application of big data technology, the research focus is placed on how to use big data for high-quality operation and precise marketing, so as to tap its potential value. Thus, the concept of "user portrait" has emerged. Today's world has completely entered the information age. The Internet provides a new perspective for human understanding of the world, and it is also a powerful weapon for human understanding of the objective world. While people enjoy the convenience and efficiency of the Internet, they are also constrained by the problems of excessive information, repeated information, and false information caused by the continuous development of the Internet, especially the continuous increase of information. In this environment, users can conveniently and accurately obtain the information they need, which becomes a problem to be solved. Users have higher demands for personal information. Therefore, on the basis of applying a recommendation algorithm, personalized recommendation for users is an inevitable development trend. Collaborative filtering is a classic technology in the field of recommendation systems, but it needs a large amount of data as support, and often encounters data sparseness in actual scenarios. In the implementation process of the technical solution of the application, the inventor found that the analytic hierarchy process (AHP) is a systematic and hierarchical multi-objective comprehensive evaluation method. In the case of complex and diverse evaluation attributes of the evaluation object, different structures, and difficulty in quantification, the AHP can also play a role. The AHP weight and the trust degree model can alleviate the sparseness of data. The setting of the AHP weight first gives each training project a weight size, so that the training project has a weight in the entire training system. In addition, in the calculation of the user trust degree, a good trust degree calculation model can improve the recommendation effect. Therefore, it is necessary to combine the AHP weight to propose an efficient and accurate recommendation scheme. SUMMARY

[0003] The application discloses a recommendation method and system based on an AHP weight and a trust degree model, to realize an efficient user personalized recommendation scheme for training project recommendation.

[0004] In one aspect, the application provides a recommendation method based on an AHP weight and a trust degree model, which comprises:

[0005] The analytic hierarchy process (AHP) is used to set the weight of each training item related to the user's ability, wherein each training item is mapped to a unique special ability index, and each special ability index corresponds to several training items;

[0006] The user similarity between each user in the database is calculated based on the trust degree model, and the similar users of each user are obtained;

[0007] The number N of recommended training items set by the target user and the target level of each special ability index are obtained, and based on the mapping relationship between the target level and the evaluation score, the target score of each training item corresponding to the current special ability index of the target user is obtained according to the score corresponding to the target level of the special ability index;

[0008] The current score of each training item of the target user is obtained: if the current training item has the evaluation score of the target user, it is taken as the current score of the current training item; otherwise, it is detected whether the similar user of the target user has the evaluation score of the current training item, if the similar user exists, the evaluation score of the similar user is taken as the current score of the current training item, otherwise the current score of the current training item is zero;

[0009] The score difference of each training item of the target user is obtained based on the current score and the target score, and then multiplied by the weight of each training item to obtain the recommendation degree of each training item;

[0010] The training item recommendation list of the target user is generated based on the top N training items with the largest recommendation degree, and is pushed to the target user.

[0011] Further, when the training item recommendation list is pushed to the target user, the recommended training time of each recommended item in the training item recommendation list is also pushed to the target user;

[0012] The calculation method of the recommended training time of the recommended item is:

[0013] The ratio of the recommendation degree of each recommended item to the sum of the recommendation degrees of all recommended items in the training item recommendation list is taken as the difference weight of the current recommended item, and the recommended training time of the current recommended item is obtained based on the product of the difference weight of the recommended item and the total training time of all recommended items.

[0014] Further, when the weights of each training item related to the user's ability are set by the analytic hierarchy process (AHP), the user's ability is divided into four levels, which are:

[0015] The comprehensive ability layer refers to the overall ability evaluation of the evaluated user, which is the weighted comprehensive result of each special ability;

[0016] The special ability layer refers to several special ability indexes of the user to be evaluated, preferably, the special ability indexes include professional ability, work ability and research ability;

[0017] The basic ability layer refers to basic index items included in each special ability index of the user to be evaluated,

[0018] The training project layer refers to several training projects included in each basic index item of the user to be evaluated.

[0019] Preferably, the basic index items corresponding to the professional ability, the work ability and the research ability are as follows:

[0020] Professional ability: professional basic knowledge, actual operation and use;

[0021] Work ability: planning and design, organization and planning, situation disposal, network management;

[0022] Research ability: theoretical research, technological innovation, subject research.

[0023] Further, the trust degree model is specifically as follows:

[0024]

[0025] Wherein, TR u,v represents the similarity between the user u and the trusted user v, STD u,v represents the subjective trust degree between the user u and the trusted user v, ID v represents the influence degree of the trusted user v, STD u,v and ID v The specific calculation formula of and ID

[0026]

[0027] Wherein, num represents the number of training projects rated by the user u and the trusted user v (both have evaluation results), nv represents the number of training projects rated by the trusted user v (has evaluation results), nu represents the number of projects rated by the user u, P v represents the number of trusted users of the trusted user v, P min represents the minimum number of trusted users in the specified trust data set, P max represents the maximum number of trusted users in the specified trust data set, r i represents the average rating of the project i, r v,i represents the rating of the project i by the trusted user v, the trusted user v ∈ V, V represents the specified trusted user set.

[0028] It should be noted that the value of P v may be based on the subjective trust degree STDu,v determination, for example, when the value of STD u,v When the value of STD is greater than a specified subjective trust threshold, the current user is considered to be a trustor of the trust user v. The extreme value of the number of trust users in the trust data set can be considered a constant, i.e., when the corresponding trust data set is specified from the target data set (which can be updated in real time with the generation of the recommendation list (training item recommendation list) and training feedback, and the original target data set is the user historical behavior data collected in a certain period of time), the upper and lower limit values of the possible number of trust users can be determined; the trust user set V can be updated in real time with the recommendation, and its initial value can be set as specified, i.e., within the extreme range of the number of trust users in the specified trust data set, the initial trust user is set. At the same time, with the generation of the recommendation list and the corresponding training feedback, the user who meets the specified conditions such as the number of training items, the average evaluation score of the training items and / or the comment information of the training items can be added to the trust user set V as a new trust user.

[0029] Further, the method of the present application further comprises: after obtaining the number of training items N of the target user and the target level of each special ability index, first screening the training items corresponding to each special ability index according to the user identity category, screening out the training items matching the target user identity category, and then determining the target score and the current score of each training item screened out, thereby realizing personalized training item recommendation according to the user category.

[0030] On the other hand, the present application also provides a recommendation system based on AHP weight and trust model, which comprises a user interaction module, a data acquisition module, a recommendation processing module and a data storage module;

[0031] The user interaction module is used to realize data interaction with the user, including: user identity login, target user input of the number of recommended training items N and the target level of each special ability index; and visual display of the generated training item recommendation list to the target user;

[0032] When the user logs in based on the user interaction module, the user interaction module matches the currently input user identity information with the user identity information in the user information table in the data storage module, and if the matching is successful, the login is allowed, otherwise it is rejected;

[0033] When the user inputs the number of items N and the target level of each special ability index based on the user interaction module, the user interaction module sends it to the recommendation processing module;

[0034] The data collection module is configured to collect historical behavior data (such as browsing and operation behavior of the user) of the user to form log records and store the log records in log files of the data storage module, and collect user basic data (id, education, name), a training project recommendation list of the user, and evaluation results of training projects of the user, and store the collected information in a designated storage format to a designated position (i.e., into corresponding forms) of the data storage module, for example, storing the user basic data into a user information table, storing the training project recommendation list into a user project plan table, and storing the evaluation results of the training projects into a user evaluation table.

[0035] The recommendation processing module is configured to generate a training project recommendation list of the target user and feed back to the user interaction module to visually display the training project recommendation list to the target user.

[0036] The recommendation processing module is configured to generate a training project recommendation list of the target user and feed back to the user interaction module to visually display the training project recommendation list to the target user.

[0037] The recommendation processing module is configured to generate a training project recommendation list of the target user and feed back to the user interaction module to visually display the training project recommendation list to the target user.

[0038] The recommendation processing module is configured to generate a training project recommendation list of the target user and feed back to the user interaction module to visually display the training project recommendation list to the target user.

[0039] The recommendation processing module is configured to generate a training project recommendation list of the target user and feed back to the user interaction module to visually display the training project recommendation list to the target user.

[0040] The recommendation processing module is configured to generate a training project recommendation list of the target user and feed back to the user interaction module to visually display the training project recommendation list to the target user.

[0041] The recommendation processing module is configured to generate a training project recommendation list of the target user and feed back to the user interaction module to visually display the training project recommendation list to the target user.

[0042] The recommendation processing module is configured to generate a training project recommendation list of the target user and feed back to the user interaction module to visually display the training project recommendation list to the target user.

[0043] Further, the recommendation module further comprises: for the currently logged-in user, reading the user basic information and recent evaluation results of the user from the data storage module, forming a user capability portrait, and visually displaying the user capability portrait through the user interaction module.

[0044] Further, the user interaction module is further used for responding to the operation feedback of the user to each recommended item in the recommended item list of the user, and sending the operation feedback to the recommendation module;

[0045] The recommendation module obtains the examination results of the user to the current recommended items based on the preset examination rules and stores the examination results into the data storage module, and the recommendation module obtains the evaluation results of the corresponding evaluation indexes based on the weights of the recommended items determined by the AHP and the user ability levels and their weights, and presents the evaluation results in a graphical form through the user interaction module.

[0046] The technical solution provided by the present application at least brings the following beneficial effects:

[0047] The present application analyzes the weak subjects that have greater influence on the user by AHP weighting the score difference of the training items, so as to effectively improve the pertinence of the recommended training items, and the similarity of the user is calculated by the trust model to obtain similar users of the user, and finally the collaborative filtering priority recommendation is performed, so that the recommended mode can effectively alleviate the data sparsity problem in the process of the collaborative filtering priority recommendation, and effectively improves the efficiency and accuracy of the collaborative filtering recommendation. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0049] Figure 1 is the AHP hierarchical structure diagram of the user ability in the embodiment of the present application;

[0050] Figure 2 is the system framework diagram of the recommendation system based on the AHP weight and the trust model provided by the embodiment of the present application;

[0051] Figure 3 is the work flow diagram of the recommendation system based on the AHP weight and the trust model provided by the embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the present application more clear, the embodiments of the present application will be further described in detail below with reference to the drawings.

[0053] Personalized recommendation is a recommendation method and system that makes recommendations tailored to each user's characteristics. This invention provides a recommendation method and system based on AHP weights and trust models to achieve accurate personalized recommendations for users.

[0054] As one possible implementation, in this embodiment of the invention, when using AHP to divide the user capability structure into several layers, this embodiment divides the user capability structure into four layers, such as... Figure 1 As shown, it includes: a comprehensive ability layer, a specialized ability layer (first-level indicators), a basic ability layer (second-level indicators), and a training program layer (third-level indicators). The comprehensive ability layer refers to the overall ability evaluation of the user being assessed, which is a weighted comprehensive result of each specialized ability; the specialized ability layer refers to the user's professional ability, work ability, and technical research ability, which is the result of a correlation analysis of each basic ability; the basic ability layer refers to the basic indicator items included in each specialized ability indicator of the user being assessed; and the training program layer refers to the specific training programs included in the basic indicator items of the assessed object, with each training program assigned a corresponding score (assessment score) through evaluation.

[0055] Next, the AHP (Aspect-Based Hierarchy) weight values ​​for user capabilities are determined. This is achieved using both hierarchical single-level ranking (ranking the importance of factors within a given level relative to a higher level) and hierarchical overall ranking (ranking the relative importance of all factors within a level to the overall goal), combined with domain expert evaluation recommendations. The specific weight values ​​for each AHP level are shown in Table 1.

[0056] Table 1. Weighting values ​​of users' comprehensive abilities

[0057]

[0058]

[0059] In this embodiment of the invention, the pre-set training items mainly include: basic principle knowledge, professional system knowledge, software system operation, hardware system operation, planning and design - requirements analysis, planning and design - scheme formulation, organization and planning - requirements analysis, organization and planning - scheme formulation, organization and planning - organization and implementation, situation handling - situation analysis, situation handling - scheme formulation, situation handling - organization and implementation, situation handling - analysis and evaluation, network management - resource management, network management - organization optimization, network management - operation analysis, theoretical research - academic papers, theoretical research - standards and specifications, technological innovation - cutting-edge technology, technological innovation - innovation and transformation, research project - scientific research project, research project - special topic research.

[0060] When the similarity users of each user in the background user information table (a user database used to record user basic information, which can generally include name, category, job level, education, learning major, professional, etc.) are obtained based on the similarity model, the target user sets the number N of recommended training courses selected, and the training object plans the target level of each special ability index. The training project recommendation list of the target user can be generated according to the following process:

[0061] (1) Calculate the weighted difference value of user ability training:

[0062] In this embodiment, the levels of the three special abilities of working ability, technical research ability and professional ability that need to be reached are set to five levels: A, B, C, D and E levels, wherein the specific division standards of each level are: A is 90-100 points, B is 80-90 points, C is 70-80 points, D is 60-70 points, and E is below 60 points.

[0063] Based on the configured score mapping rule, the three levels of working ability, technical research ability and professional ability input by the target user are mapped to the corresponding target score CS. For example, the middle value of each segment is taken as the mapping result of the level, or different identities of the user can be configured based on the identity category, and the lower limit or upper limit of the value range of the level is mapped, etc.

[0064] In the embodiment of the application, the corresponding training projects are matched for users of different identity categories, that is, the training projects are first primarily screened based on the identity category of the target user, and then the final recommended projects are obtained from the primarily screened projects to push to the target user.

[0065] In the primarily screened training projects, the current score PS of each training project is determined: if the score of the target user on the current training project i is recorded, it is directly taken as the current score Psi; otherwise, the score of the similar user of the target user on the training project i is set based on the score of the similar user of the target user on the training project i: if the score of the similar user of the target user on the training project i is recorded, it is directly taken as the current score Psi, and if no record is recorded, the current score Psi is directly set to 0.

[0066] That is, by comparing with the training evaluation scores already obtained by the user, the training courses that need to be improved are found out, and the difference D between the current score PS and the target score CS of each training project is calculated, that is:

[0067] Di=PSi-CSi

[0068] Wherein, i is the i-th training project, PSi is the current score of the training project i, and CSi is the target score of the training project i.

[0069] The size of the difference Di reflects the distance of the target user from the target result on the training item i. In the embodiment of the present application, the recommendation is not completely based on the size of the difference, because each training item has its weight, and the size of the training item weight means the importance of the training item in the training process. Therefore, in the recommendation process, not only the difference but also the size of the weight is considered. Therefore, the product R of the difference D and the weight W is used to determine the recommendation sequence, that is:

[0070] Ri=Di×Wi

[0071] where i is the i-th training item, Di is the result difference of the training item i, and Wi is the weight value of the training item i, which can be determined based on Table 1.

[0072] (2) Personalized training plan recommendation for the user.

[0073] The product Ri of the difference and the weight of each training item, the larger the value of Ri, the more the training item i should be arranged in the front of the recommendation plan sequence, and vice versa, the smaller the value of Ri, the more the training item i should be arranged in the back of the recommendation plan sequence. The TOP-N recommendation is used to generate the recommendation sequence, and the first N of the recommendation sequence is taken to obtain the recommended training plan.

[0074] (3) Calculation of the training time length of the recommended training item.

[0075] In the embodiment, the algorithm of the product of the non-fixed time length and the difference weight ratio is used. The non-fixed time length means that the total training time length can be flexibly set. Because different training items, training scenes, training intensities and other factors will affect the training time length, in the example of the present application, the time length is set as a variable parameter, and the specific calculation formula of the difference weight ratio is as follows:

[0076]

[0077] where Qi is the difference weight ratio, which means that the product (Ri) of the result difference and the weight of each recommended training item i in the generated recommended training plan accounts for the proportion of the sum of the difference weight of all recommended items.

[0078] Hi=Qi×Th

[0079] where Hi is the recommended training time length of the training item i, and Th is the total training time length, which is a preset value.

[0080] As a possible implementation manner, as shown in Figure 2 The recommendation system based on the AHP weight and trust degree model provided by the embodiment of the present application includes a user interaction module, a data acquisition module, a recommendation processing module and a data storage module.

[0081] Among them, the user uses the inter-module to realize the user login, the input of the recommended parameter information (the number of projects, the target level of the special ability index, etc.), the operation feedback of the training project, and other user input behaviors, and the graphical method presents the processing results of the recommended processing module to the user, etc.

[0082] The data collection module is used to collect the historical behavior data of the user to form a log record and store it in the log file of the data storage module. The behavior data of the user is one of the important data sources of the recommendation system, so the historical behavior data of the user needs to be collected through the log collection layer, and these data are stored in the log file. According to the different characteristics of the behavior, the corresponding information is extracted to generate a log record, such as for browsing data, extracting user identification, browsing identification and browsing timestamp to form a log record. The original log file is stored in a certain format and belongs to semi-structured data, which may contain some redundant data, so it is necessary to clean, delete or supplement the incomplete data in the log, remove the redundant data, and realize the normalization of the log.

[0083] Meanwhile, the data collection module is also used for collecting user basic data (id, education, name), user training project recommendation list and user training project evaluation results, etc., and storing the collected corresponding information into the corresponding forms of the data storage module according to the specified storage format, preferably, in the embodiment of the present application, the user information table (user), the plan table (plan) for storing the training project recommendation list, the training content table (splan), the project index table (qzzb) and the project evaluation result table (nlxm) are arranged in the data storage module. That is, in the embodiment, the user information is stored in the user table, the table contains fields of userid (user id), education, name, user category, user ability, etc., wherein the userid is set to be unique, the field names zhnldj (comprehensive ability level), zynldj (professional ability level), gznldj (work ability level) and jydndj (technical research ability level) in the user ability are automatically calculated after the user fills in the evaluation results, and are obtained by layer-by-layer calculation combined with the evaluation results and the weight shown in Table 1. The plan table mainly stores the user's plan, the table is associated to the user table through the foreign key userid, so that the user corresponds to the user's plan, and the plan table is also provided with a reserved expansion field (Tjnxlm). The qzzb table is mainly used for storing the training project indexes obtained by the user through the AHP model, and matching the training project indexes corresponding to different user categories based on the user category identification field (lb). The nlxm table mainly stores the evaluation results of the corresponding ability projects of the user, wherein the userid is associated to the user table through the foreign key, so that the detailed evaluation results corresponding to the user can be queried through the userid. The splan table is used for storing the specific content of the user's training plan, which includes the ability projects and the training duration, and a difference mark field of the ability level set by the user and the current level.

[0084] The recommendation processing module is used for constructing a user portrait model, generating a training project recommendation list (personalized recommendation scheme) of a target user, and feeding back the user portrait and the personalized recommendation scheme to the user interaction module, so as to visually output and display to the target user.

[0085] The recommendation processing module is based on user historical behavior data in the log file when constructing a user portrait model to generate a user portrait of the current user, and then constructs the user portrait model through analysis and processing of user attribute information (user basic information) and historical behavior (such as recent evaluation results). For example, the output user portrait information includes: user basic information (name, job level, education, professional, learning professional, gender, category, etc.). The evaluation results of the AHP user ability layer show that the evaluation results of each level index obtained based on the evaluation results and the AHP weight of each layer, for example, for each training project of the training project layer, the evaluation result is directly the evaluation result; for each index of the upper layer (basic ability layer) of the training project, the evaluation result of each basic index is: the weighted average of the evaluation results of all training projects under the current basic index and the weight; and the evaluation result of each special ability index of the special ability layer is: the weighted average of the evaluation results of all basic indexes under the current special ability index. Finally, the evaluation result of the comprehensive ability is obtained based on the weighted average of the evaluation results of all special ability indexes.

[0086] In addition, when the user implements the current personalized recommendation scheme based on the user interaction module, the data acquisition module also updates the current evaluation result to the data storage module in real time, so that the recommendation processing module can generate a real-time personalized recommendation scheme based on the latest user behavior data when generating the personalized recommendation scheme of the current target user next time. At the same time, for the training user whose training evaluation result does not reach the set ability requirement, a new personalized recommendation scheme is generated according to the current training evaluation result, and a new round of training process is carried out again.

[0087] As shown in Figure 3 The working process of the AHP weight and trust degree model-based recommendation system provided by the embodiment of the application through a visual interface mainly includes:

[0088] (1) Open the browser to enter the system login page and complete the administrator user login;

[0089] (2) Collect basic data and evaluation data, use the data acquisition module to collect the basic information of the training user and the recent evaluation data, form a user ability portrait, and reflect the basic attributes, ability attributes and other information of the training user;

[0090] (3) Plan scheme development, use the recommendation processing module, select the "ability requirement" of the plan according to the ability portrait of the training user, that is, the ability level that the training user plans to reach in the comprehensive ability, professional ability, work ability and technical research ability, and generate a personalized scheme matched with the "ability requirement" of the training user through the system recommendation algorithm.

[0091] (4) Training effect evaluation, using the recommendation processing module, according to the score of each training project after the implementation of the formulated plan scheme, through the AHP evaluation algorithm, the evaluation values of the comprehensive ability, professional ability, working ability and technical research ability of the training user are calculated, and the evaluation analysis results of the ability scores and effects of the training user at each level are presented in a graphical manner.

[0092] (5) Training data collection, using the data collection module, the training evaluation score data of the training user after the implementation of the training plan is collected and entered, which is used as the data basis for the generation of a new plan scheme.

[0093] (6) Plan scheme adjustment, using the recommendation processing module, for the training user whose training evaluation score does not reach the set ability requirement, according to the current training evaluation score, the training plan scheme is adjusted through the recommendation algorithm, and a new round of training process is carried out again.

[0094] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

[0095] The above only describes some embodiments of the present application. For those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the protection scope of the present application.

Claims

1. A recommendation method based on AHP weight and trust degree model, characterized in that, The method comprises the following steps: An analytic hierarchy process (AHP) is used to set the weight of each training item related to the user's ability, wherein each training item is mapped to a unique special ability index, and each special ability index corresponds to several training items; A trust degree model is used to calculate the user similarity between users in the database, and obtain the similar users of each user; The number N of recommended training items set by the target user and the target level of each special ability index are obtained, and based on the mapping relationship between the target level and the evaluation score, the target score of each training item corresponding to the current special ability index of the target user is obtained according to the score corresponding to the target level of the special ability index; The current score of each training item of the target user is obtained: if the target user has an evaluation score for the current training item, the evaluation score is taken as the current score of the current training item; otherwise, it is detected whether the similar user has an evaluation score for the current training item, if the similar user has, the evaluation score is taken as the current score of the current training item, otherwise the current score of the current training item is set to zero; The score difference of each training item of the target user is obtained based on the current score and the target score, and then multiplied by the weight of each training item to obtain the recommendation degree of each training item; The training item recommendation list of the target user is generated based on the top N training items with the largest recommendation degree, and pushed to the target user; The trust degree model is specifically: wherein, represents the similarity between the user u and the trust user v, represents the subjective trust degree between the user u and the trust user v, represents the influence degree of the trust user v, and The specific calculation formula is: ， ， wherein, represents the number of training items co-rated by user u and trusted user v, represents the number of training items rated by trusted user v, represents the number of items rated by user u, represents the number of trusted users of trusted user v, represents the minimum number of trusted users in the specified trust data set, represents the maximum number of trusted users in the specified trust data set, represents the average rating of item i, represents the rating of item i by trusted user v, trusted user , V represents the specified set of trusted users.

2. The method of claim 1, wherein, When the training item recommendation list is pushed to the target user, the recommended training time of each recommended item in the training item recommendation list is also pushed to the target user; The calculation method of the recommended training time of the recommended item is: The ratio of the recommendation degree of each recommended item to the sum of the recommendation degrees of all recommended items in the training item recommendation list is taken as the difference weight of the current recommended item, and the recommended training time of the current recommended item is obtained based on the product of the difference weight of the recommended item and the total training time of all recommended items.

3. The method of claim 1, wherein, When the analytic hierarchy process (AHP) is used to set the weight of each training item related to the user's ability, the user's ability is divided into four levels, which are: The comprehensive ability layer refers to the overall ability evaluation of the evaluated user, which is the weighted comprehensive result of each special ability; The special ability layer refers to several special ability indexes of the evaluated user; The basic ability layer refers to the basic index items included in each special ability index of the evaluated user; The training item layer refers to several training items included in each basic index item of the evaluated user.

4. The method of claim 3, wherein, The special ability index includes professional ability, work ability and technical research ability; The professional ability includes professional basic knowledge and practical operation; The work ability includes planning and design, organization and planning, situation handling and network management; The technical research ability includes theoretical research, technical innovation and project research. ​ 5. The method of claim 1, wherein, After obtaining the number N of training projects of the target user and the target level of each special ability index, first, according to the user identity category, the training projects corresponding to each special ability index are initially screened, and the training projects matching the target user identity category are screened out, and then the target score and the current score of each training project screened out are determined, so as to realize the personalized training project recommendation for the user category.

6. A recommendation system based on AHP weight and trust degree model, for implementing the recommendation method of any one of claims 1-5, characterized in that, The system comprises a user interaction module, a data collection module, a recommendation processing module and a data storage module. The user interaction module is used for realizing data interaction with the user, including: user identity login, target user input of the number N of recommended training projects and the target level of each special ability index; and visual display of the generated training project recommendation list to the target user. When the user logs in based on the user interaction module, the user interaction module matches the current input user identity information with the user identity information in the user information table in the data storage module, and if the matching is successful, the login is allowed, otherwise it is rejected. When the user inputs the number N of projects and the target level of each special ability index based on the user interaction module, the user interaction module sends them to the recommendation processing module. The data collection module is used for collecting the historical behavior data of the user to form a log record and storing it in the data storage module, and collecting the user basic data, the training project recommendation list of the user and the evaluation score of the training project of the user, and storing the collected corresponding information in the specified location of the data storage module according to the specified storage format. The recommendation processing module is used for generating the training project recommendation list of the target user and feeding back to the user interaction module, so as to realize visual display to the target user. The generation of the training project recommendation list of the target user is specifically as follows: Based on the preset trust degree model, the user similarity between each user in the user information table of the data storage module is calculated to obtain the similar users of each user. The current input number N of projects and the target level of each special ability index are taken as the target user. The mapping relationship between the target level and the evaluation score is read from the data storage module, and the target score of each training project corresponding to the current special ability index of the target user is obtained based on the score corresponding to the target level of the special ability index. Based on the evaluation score of the training project stored in the data storage module, the current score of each training project of the target user is obtained: if the target user has the evaluation score of the current training project, it is taken as the current score of the current training project; otherwise, it is detected whether the similar user of the target user has the evaluation score of the current training project, if the similar user exists, the evaluation score is taken as the current score of the current training project, otherwise the current score of the current training project is zero. Based on the current score and the target score, the score difference of each training project of the target user is obtained, and then it is multiplied by the weight of each training project to obtain the recommendation degree of each training project. The training project recommendation list of the target user is generated based on the first N training projects with the largest recommendation degree and is pushed to the target user.

7. The recommendation system of claim 6, wherein, The recommendation processing module, when generating the training item recommendation list, further comprises a recommended training duration of each recommended item; The recommended training duration of each recommended item is: taking a ratio of a recommendation degree of each recommended item to a sum of recommendation degrees of all recommended items in the training item recommendation list as a difference weight of the current recommended item, and obtaining a recommended training duration of the current recommended item based on a product of the difference weight of the recommended item and a total training duration of all recommended items.

8. The recommendation system of claim 6 or 7, wherein, The recommendation processing module further comprises: reading, from the data storage module, user basic information and a recent evaluation result of a currently logged-in user, forming a user capability profile, and visually displaying the user capability profile through the user interaction module.

9. The recommendation system of claim 6 or 7, wherein, The user interaction module is further configured to respond to operation feedback of the user on each recommended item in the training item recommendation list of the user and send the operation feedback to the recommendation processing module; The recommendation processing module obtains an evaluation result of the user on each recommended item based on a preset evaluation rule and stores the evaluation result in the data storage module, and the recommendation processing module obtains an evaluation result of a corresponding evaluation index layer by layer based on the weight of each recommended item determined by the AHP, the user capability hierarchy and the weight of each user capability hierarchy, and presents the evaluation result in a graphical form through the user interaction module.

Images