Personalized information resource recommendation method and device and computer equipment
Through the information resource recommendation model of multi-source data acquisition and real-time update, the problem of insufficient accuracy and timeliness of information resource recommendation in existing systems is solved, and higher user satisfaction is achieved.
Patent Information
- Application Number
- CN202411877099.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-19
- Publication Date
- 2025-04-11
AI Technical Summary
The existing personalized information resource recommendation system has shortcomings in the accuracy and timeliness of recommendation results, resulting in a decrease in user satisfaction.
By obtaining the target user's multi-source historical data and real-time data, the information resource recommendation model is built and updated to generate more accurate and timely information resource recommendation results.
It improves the accuracy and timeliness of the information resource recommendation results, thereby improving user satisfaction.
Smart Images
Figure CN119807528A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing technology, and in particular to a personalized information resource recommendation method, device and computer equipment. Background Art
[0002] In recent years, with the explosive growth of Internet information, users often face the problem of information overload when obtaining the network information they need. In order to solve the problem of information overload when users obtain information, personalized information resource recommendation systems have emerged. Personalized information resource recommendation systems analyze users' historical behavior data and interest preferences to provide users with personalized information resource recommendations to improve user satisfaction.
[0003] In related technologies, personalized information resource recommendation systems mainly rely on data such as users' historical behaviors and interest preferences. Through algorithms such as content filtering and collaborative filtering, and based on the similarity between the content features of information resources and users' historical behaviors and interest preferences, they predict information resources that users may be interested in and recommend them to users.
[0004] However, when recommending information resources to users based on relevant technologies, the source of the user's historical behavior data and interest preference data is single, and the user's historical behavior data and interest preference data are relatively sparse, which leads to poor accuracy of information resource recommendation results. In addition, when recommending information resources based on the user's historical behavior data and interest preference data, it is not possible to timely discover changes in user's interest preferences and the emergence of new information resources, which leads to poor timeliness of information resource recommendation results and reduces user satisfaction. Summary of the invention
[0005] The purpose of this application is to provide a personalized information resource recommendation method, device and computer equipment, which can improve the accuracy and timeliness of information resource recommendation results to improve user satisfaction.
[0006] The embodiment of the present application is implemented as follows:
[0007] In a first aspect of an embodiment of the present application, a personalized information resource recommendation method is provided, the method comprising:
[0008] Obtain historical data of target users, including historical behavior data, historical social data, user portrait data, historical comment data, and historical feedback data;
[0009] Building an information resource recommendation model based on historical data of the target user, and generating an information resource recommendation result by the information resource recommendation model, wherein the information resource recommendation result includes at least one information resource file;
[0010] Acquire the real-time data of each information resource file in the information resource recommendation result of the target user, and update the information resource recommendation model according to the real-time data to obtain the target information resource recommendation model. The real-time data includes: current behavior data, current social data, current comment data and current feedback data;
[0011] The target information resource model generates and outputs the target information resource recommendation results.
[0012] As a possible implementation method, an information resource recommendation model is constructed based on the historical data of the target users, including:
[0013] Perform data cleaning and data preprocessing on the historical data of target users to obtain valid historical data;
[0014] Build an information resource recommendation model based on valid historical data.
[0015] As a possible implementation method, an information resource recommendation model is constructed based on valid historical data, including:
[0016] Divide the valid historical data into training data set and test data set;
[0017] An initial information resource recommendation model is obtained through training and learning based on the training data set, and the initial information resource recommendation model is tested through the test data set;
[0018] According to the test results of the initial information resource recommendation model, determine the model evaluation indicators corresponding to the initial information resource recommendation model, and the model evaluation indicators include: accuracy, recall rate and neutralization value;
[0019] The model parameters of the initial information resource recommendation model are adjusted according to the model evaluation indicators, and the user preference characteristics are determined to obtain the information resource recommendation model. The model parameters include: learning rate, batch size and number of hidden layers.
[0020] As a possible implementation method, the information resource recommendation model generates information resource recommendation results, including:
[0021] The information resource recommendation model determines the similarity between the user preference feature and the content feature corresponding to each information resource file according to the user preference feature and the content feature of each information resource file in the information resource database;
[0022] According to the order of similarity, information resource recommendation results are generated.
[0023] As a possible implementation method, the information resource recommendation model is updated according to the real-time data to obtain the target information resource recommendation model, including:
[0024] Determine whether each information resource file is a valid information resource file according to the real-time data of each information resource file in the information resource recommendation result by the target user;
[0025] If so, the user preference features in the information resource recommendation model are updated according to the content features of the valid information resource files.
[0026] As a possible implementation method, determining whether each information resource file is a valid information resource file according to the real-time data of each information resource file in the information resource recommendation result by the target user includes:
[0027] If the target user views the current information resource file in the information resource recommendation result, and the viewing time of the current information resource file reaches the preset time threshold, the current information resource file is a valid information resource file;
[0028] If the target user has not viewed the current information resource file in the information resource recommendation results, and / or the viewing time of the current information resource file does not reach the preset time threshold, the current information resource file is an invalid information resource file and is deleted from the information resource recommendation results.
[0029] As a possible implementation method, the above-mentioned personalized information resource recommendation method further includes:
[0030] Filter the historical data of the target user based on a preset time period to obtain the historical data of the target user within the target period;
[0031] Extract features from the historical data of the target user in the target period to obtain the user preference features in the target period;
[0032] The information resource recommendation model is updated in real time based on user preference characteristics within the target period.
[0033] As a possible implementation method, the above-mentioned personalized information resource recommendation method further includes:
[0034] Update the weight value of the user preference feature according to the preset forgetting function and the historical data of the target user;
[0035] The information resource recommendation model is updated in real time according to the updated weight values of user preference features.
[0036] According to a second aspect of an embodiment of the present application, a personalized information resource recommendation device is provided, the device comprising:
[0037] The acquisition module is used to obtain the historical data of the target user, including historical behavior data, historical social data, user portrait data, historical comment data and historical feedback data;
[0038] A construction module, used to construct an information resource recommendation model based on the historical data of the target user, and generate an information resource recommendation result by the information resource recommendation model, wherein the information resource recommendation result includes at least one information resource file;
[0039] An updating module is used to obtain the real-time data of each information resource file in the information resource recommendation result of the target user, and update the information resource recommendation model according to the real-time data to obtain the target information resource recommendation model. The real-time data includes: current behavior data, current social data, current comment data and current feedback data;
[0040] The generation module is used to generate and output the target information resource recommendation results based on the target information resource model.
[0041] According to a third aspect of an embodiment of the present application, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the personalized information resource recommendation method described in the first aspect is implemented.
[0042] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the personalized information resource recommendation method described in the first aspect is implemented.
[0043] The beneficial effects of the embodiments of the present application include:
[0044] The embodiment of the present application provides a personalized information resource recommendation method, which obtains the historical data of the target user on each information resource file in the information resource database, and obtains the information resource recommendation model through training and learning based on the historical data of the target user, and generates the information resource recommendation result through the information resource recommendation model; obtains the real-time data of the target user on each information resource file included in the information resource recommendation result, determines whether the current interest preference of the target user has changed compared with the historical interest preference, and updates the information resource recommendation model based on the real-time data of the target user on each information resource file in the information resource recommendation result, so as to obtain the target information resource recommendation model adapted to the current interest preference of the target user, and generates the target information resource recommendation result through the target information resource recommendation model. Among them, the historical data of the target user includes historical data from multiple data sources such as historical behavior data, historical social data, user portrait data, historical comment data, and historical feedback data, so that the sample data for training and learning the information resource recommendation model can be increased, thereby improving the accuracy and stability of the information resource recommendation model. In addition, based on the target user's real-time data on each information resource file in the information resource recommendation result, the target user's interest preference changes are predicted, and then the information resource recommendation model is updated in real time, so that the target information resource recommendation model matches the target user's current interest preference in real time, which can improve the timeliness of the information resource recommendation result. In this way, the accuracy and timeliness of the information resource recommendation result can be improved to improve the user's satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0046] Figure 1 A flowchart of a first personalized information resource recommendation method provided in an embodiment of the present application;
[0047] Figure 2 A flowchart of a second personalized information resource recommendation method provided in an embodiment of the present application;
[0048] Figure 3 A flowchart of a third personalized information resource recommendation method provided in an embodiment of the present application;
[0049] Figure 4 A flowchart of a fourth personalized information resource recommendation method provided in an embodiment of the present application;
[0050] Figure 5A flowchart of a fifth personalized information resource recommendation method provided in an embodiment of the present application;
[0051] Figure 6 A flowchart of a sixth personalized information resource recommendation method provided in an embodiment of the present application;
[0052] Figure 7 A flowchart of a seventh personalized information resource recommendation method provided in an embodiment of the present application;
[0053] Figure 8 A flowchart of an eighth personalized information resource recommendation method provided in an embodiment of the present application;
[0054] Fig. 9 A flowchart of a complete personalized information resource recommendation method provided in an embodiment of the present application;
[0055] Fig.10 A flow chart of a method for updating configuration item information provided in an embodiment of the present application;
[0056] Fig.11 A flow chart of a method for determining a target configuration item to be updated provided in an embodiment of the present application;
[0057] Fig.12 A schematic diagram of the structure of a personalized information resource recommendation device provided in an embodiment of the present application;
[0058] Fig.13 A schematic diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0059] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.
[0060] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.
[0061] At present, personalized information resource recommendation systems mainly rely on data such as users' historical behaviors and interest preferences. Through algorithms such as content filtering and collaborative filtering, and based on the similarity between the content characteristics of information resources and users' historical behaviors and interest preferences, they predict information resources that users may be interested in and recommend them to users. However, this solution has the problem that the source of users' historical behavior data and interest preference data is single, and the historical behavior data and interest preference data of users are relatively sparse, which leads to poor accuracy of information resource recommendation results. In addition, this solution cannot timely detect changes in users' interest preferences and the emergence of new information resources, which leads to poor timeliness of information resource recommendation results, resulting in reduced user satisfaction.
[0062] To this end, the embodiment of the present application provides a personalized information resource recommendation method, which collects historical data of target users through multiple data sources, and constructs an information resource recommendation model based on the historical data of the target users. The information resource recommendation model generates information resource recommendation results based on the content characteristics of each information resource file in the information resource database; obtains the real-time data of the target user on each information resource file in the information resource recommendation result, updates the information resource recommendation model to obtain the target information resource recommendation model, and generates the target information resource recommendation result through the target information resource recommendation model. In this way, the accuracy and timeliness of the information resource recommendation results can be improved to improve the user's satisfaction with the use.
[0063] The personalized information resource recommendation method provided in the embodiment of the present application is explained in detail below with reference to the accompanying drawings.
[0064] Figure 1 A flowchart of a personalized information resource recommendation method provided for this application, which can be applied to an information resource recommendation system, wherein the information resource recommendation system includes a running proxy component, a virtual machine, and an external cluster, and the running proxy component is used to realize information interaction between the external cluster and the virtual machine. Among them, the running proxy component is used to act as an intermediary between the client and the server; the virtual machine is an independent and isolated complete computer system simulated by the user through software; the external cluster is composed of a group of independent computers, and the external cluster is a large-scale computer service system established based on a high-speed communication network, each cluster node is an independent server, the server is deployed on a cloud container, the server is integrated with relevant data models, and the historical data of the target user is collected through the server. See Figure 1 , the embodiment of the present application provides a personalized information resource recommendation method, comprising:
[0065] S101. Obtain historical data of the target user, where the historical data includes: historical behavior data, historical social data, user portrait data, historical comment data, and historical feedback data.
[0066] Optionally, the historical data of the target user is obtained through a server deployed on the cloud container. The target user does not refer to a single user. The target user can be used to indicate a user with the same or similar interests and hobbies. The historical data is used to indicate the target user's reference data on information resource files in the past period of time. The historical data can be the target user's reference data on information resource files in the past month, or the target user's reference data on information resource files in the past three months. This application does not make specific restrictions on this. It is worth noting that the denser the historical data sample of the target user, the more accurate the characterization of the target user's interest preferences, and the higher the accuracy of the personalized information resource recommendation results.
[0067] Optionally, the information resource file is used for various information resources such as work reports, official documents, work summary tables, etc. For example, the information resource file can be an attendance sheet for August, a pay slip for September, etc. This application does not make specific limitations on this.
[0068] Optionally, historical behavior data is used to indicate the target user's access behavior data to information resource files within a historical period. The historical behavior data may be behavior data such as browsing pages of information resource files, searching for keywords of information resource files, and clicking on information resource files for viewing; historical social data is used to indicate which social media the target user used to view information resource files within a historical period. Historical social data includes social media data such as accessing information resource files through applications and accessing information resource files through web pages; user portrait data is used to indicate the target user's feature tags. User portrait data includes data such as age and gender; historical comment data is used to indicate the target user's evaluation of information resource files within a historical period. Historical comment data may be five-star positive reviews, one-star negative reviews, etc.; historical feedback data is used to indicate the target user's rating of information resource files within a historical period.
[0069] Optionally, after deploying the log collection system on the server side, the server automatically creates a project access log file based on the log collection system. The log collection system will store the log files accessed by the target user in the specified directory, and create a corresponding folder directory according to the time of the target user's access. The log collection system can record the target user's log access behavior, including the log pages accessed by the target user, the target user's browsing time on the log page, the device information of the target user accessing the log page, etc.
[0070] Optionally, by adding a user behavior data collection function in the official document system of the server, the user behavior data collection interface is triggered by accessing the system, and then the target user's search behavior, browsing behavior, click behavior, jump behavior and other behavior data in the official document system are collected. At the same time, whether the target user clicks on the information resource files included in the current information resource recommendation results, as well as the target user's browsing behavior, browsing time, jump behavior, comment results, rating results and tags in the recommended information resource files are recorded.
[0071] Optionally, by deploying a mobile official document application collection system on the server, the mobile official document application collection system is used to collect relevant data generated when the target user uses a mobile terminal device to access the official document system, such as search keywords, browsed official document pages, reading time, comments, ratings and other data, and classify and save the relevant data generated by the information resource files accessed by the target user using the mobile terminal.
[0072] S102: construct an information resource recommendation model based on the historical data of the target user, and generate an information resource recommendation result by the information resource recommendation model, wherein the information resource recommendation result includes at least one information resource file.
[0073] Optionally, the target user's historical data is used as sample data for constructing an information resource recommendation model. The neural network model is trained and verified through the target user's historical data to obtain an information resource recommendation model that matches the target user. The information resource recommendation model includes the target user's interest configuration items. The information resource recommendation model corresponding to the target user generates an information resource recommendation result based on the matching degree between the content characteristics of each information resource file in the information resource database and the target user's interest preferences.
[0074] Optionally, the information resource recommendation model is a mathematical model trained and learned based on historical data of the target user. The information resource recommendation model generates information resource recommendation results based on the correlation between the content of each information resource file in the information resource data and the interest preferences of the target user. The information resource recommendation results are used to indicate the sequence of information resource files that meet the interest preferences of the target user. The order of each information resource file in the information resource recommendation results is determined by the configuration value of the user's interest preference and the degree of correlation between the interest preference and the content of the information resource file.
[0075] S103, obtaining the real-time data of each information resource file in the information resource recommendation result of the target user, and updating the information resource recommendation model according to the real-time data to obtain the target information resource recommendation model, the real-time data includes: current behavior data, current social data, current comment data and current feedback data.
[0076] Optionally, after the information resource recommendation model generates the information resource recommendation results, it pushes them to the device used by the target user, obtains the target user's real-time data on each information resource file in the information resource recommendation results through the server, and determines the changes in the target user's interest preferences based on the target user's real-time data on each information resource file in the information resource recommendation results.
[0077] Optionally, current behavior data is used to indicate the target user's current access information data on each information resource file in the information resource recommendation results, such as the target user's current behavior data on browsing, clicking, browsing time, jumping, etc. on each information resource file in the information resource recommendation results; current social data is used to indicate the target user's current social media access to each information resource file in the information resource recommendation results, such as the target user currently accesses each information resource file in the information resource recommendation results through an application; current comment data is used to indicate the target user's comments on each information resource file in the information resource recommendation results, such as the target user's current comments on the information resource files in the information resource recommendation results are negative, etc.; current feedback data is used to indicate the target user's rating data for each information resource file in the information resource recommendation results.
[0078] Optionally, based on the real-time data of the target user on each information resource file in the information resource recommendation result, the change of the target user's interest preference is determined, and the information resource recommendation model is updated according to the change of the target user's interest preference to obtain the target information resource recommendation model, and the target information resource recommendation model is adapted to the current interest preference of the target user. It is worth noting that the change of the target user's interest preference can be either adding a new interest preference or forgetting an old interest preference, and this application does not make specific limitations on this.
[0079] S104: Generate and output target information resource recommendation results using the target information resource model.
[0080] Optionally, a target information resource recommendation result is generated through a real-time updated target information resource model, and the target information resource recommendation result includes an information resource file sequence, wherein the target information resource recommendation result matches the current interest preference of the target user, and the information resource recommendation result is adapted to the historical interest preference of the target user.
[0081] In an embodiment of the present application, by obtaining the historical data of the target user on each information resource file in the information resource database, and training and learning the information resource recommendation model based on the historical data of the target user, the information resource recommendation result is generated through the information resource recommendation model; the real-time data of the target user on each information resource file included in the information resource recommendation result is obtained to determine whether the current interest preference of the target user has changed compared with the historical interest preference, and the information resource recommendation model is updated based on the real-time data of the target user on each information resource file in the information resource recommendation result, so as to obtain the target information resource recommendation model adapted to the current interest preference of the target user, and the target information resource recommendation result is generated through the target information resource recommendation model. Among them, the historical data of the target user includes historical data from multiple data sources such as historical behavior data, historical social data, user portrait data, historical comment data, and historical feedback data, so that the sample data for training and learning the information resource recommendation model can be increased, thereby improving the accuracy and stability of the information resource recommendation model. In addition, based on the target user's real-time data on each information resource file in the information resource recommendation result, the target user's interest preference changes are predicted, and then the information resource recommendation model is updated in real time, so that the target information resource recommendation model matches the target user's current interest preference in real time, which can improve the timeliness of the information resource recommendation result. In this way, the accuracy and timeliness of the information resource recommendation result can be improved to improve the user's satisfaction.
[0082] For a possible implementation, see Figure 2 The operation of step S102 may specifically be:
[0083] S201. Perform data cleaning and data preprocessing on the historical data of the target user to obtain valid historical data.
[0084] Optionally, after data cleaning and data preprocessing of the historical data of the target user, valid historical data is obtained, and the valid historical data is used to indicate part of the historical data of the target user that can be used for training and building the model. Among them, the data cleaning methods include: missing value processing, outlier detection and processing, duplicate data identification and processing, data format conversion, data type conversion, error value identification and processing, logical consistency check, etc.; data preprocessing methods include: data normalization processing, data discretization processing, data integration, data grouping and aggregation, data labeling, data randomization, data sharding and splitting, and data conversion, etc.
[0085] Optionally, missing value processing is performed on the historical data of the target user, missing values in the historical data of the target user are checked, and according to the actual situation, missing values are filled, missing values are deleted, or missing values are specially marked. Missing value filling often uses numerical filling such as mean, median, mode, etc.; outlier detection is performed on the historical data of the target user through statistical algorithms, visualization technology and machine learning methods, and it is decided whether to delete or correct these outliers according to the actual situation; duplicate data is identified on the historical data of the target user, and duplicate data in the historical data of the target user is deleted or merged; the historical data of the target user is converted into data in a unified format, such as unifying the date format into a year-month-day format; data type conversion is performed on the historical data of the target user so that the historical data of the target user can be accurately identified, such as converting a string into an integer or floating point number, etc.; error value identification and processing is performed on the historical data of the target user, such as converting text errors into correct values or using specific algorithms to correct errors. The error value recognition and processing specifically include: error analysis, error recognition, error handling and preventive measures. Among them, "error analysis" uses decision trees to display the characteristics of queues with higher error rates, and uses heat maps to intuitively display the impact of characteristics on the error rates of each queue. "Error value recognition" is completed through the error analysis component, which automatically discovers error queues and guides targeted mitigation steps. "Error handling" involves correcting or filling bad values or missing values. For parameter parsing errors, server internal errors, etc., they are solved by checking training parameters and data paths. "Preventive measures" can avoid common errors such as "overfitting" to the greatest extent, so as to ensure the correct configuration of the environment and quickly locate errors. Check whether the concepts, value domains, formats, topologies and other logical dimensions in the historical data of the target user are consistent, such as the degree of compliance with logical rules between the attributes of business entities such as equipment used by the target user on a daily basis and office document circulation and their data structures, compare data verification rules and their values, and find data input errors and business logic errors.
[0086] Optionally, the historical data of the target user is scaled or normalized so that the historical data of the target user falls within a certain range or is expressed on the same scale, such as normalizing the target user's age data to between 0 and 1; the historical data of the target user is converted from a continuous variable to a discrete variable to facilitate classification of the historical data of the target user or use of a decision tree algorithm; the historical data of the target user from different data sources is integrated into a unified data set to ensure the consistency and integrity of the historical data of the target user; the historical data of the target user is grouped and summarized, such as summarizing sales data by region; the classification labels in the historical data of the target user are labeled, such as labeling gender variables with 0 and 1; the historical data of the target user is randomized to ensure stability and repeatability during the training process of the information resource recommendation model; the historical data of the target user is split to obtain a training data set, a test data set, and a verification data set corresponding to the historical data of the target user; the historical data of the target user is converted into a data format suitable for a specific algorithm or model, such as converting matrix data into a data format suitable for a neural network model.
[0087] S202. Construct an information resource recommendation model based on valid historical data.
[0088] Optionally, an information resource recommendation model is obtained by training and verifying the valid historical data corresponding to the target user.
[0089] For a possible implementation, see Figure 3 The operation of step S202 may specifically be:
[0090] S301. Divide the valid historical data into a training data set and a test data set.
[0091] Optionally, call a data splitting tool to randomly split the valid historical data corresponding to the target user into a training data set and a test data set, wherein both the training data set and the test data set contain multiple valid historical data, the valid historical data contained in the training data set are used as sample data for training the initial information resource recommendation model, and the valid historical data contained in the test data set are used as test sample data for performance testing of the initial information resource recommendation model.
[0092] S302: An initial information resource recommendation model is obtained by training and learning based on the training data set, and the initial information resource recommendation model is tested by using the test data set.
[0093] Optionally, the initial information resource recommendation model is an information resource recommendation model obtained by repeated training and learning based on training sample data in the training data set, and the evaluation index of the initial information resource recommendation model has not been tested with test sample data.
[0094] Optionally, the evaluation performance of the initial information resource recommendation model is cross-validated using test sample data included in the test data set, wherein the test data included in the test data set is used to verify the degree of conformity of the training model parameters in the initial information resource recommendation model to the test data other than the training data.
[0095] S303: Determine the model evaluation index corresponding to the initial information resource recommendation model according to the test result of the initial information resource recommendation model. The model evaluation index includes: accuracy, recall rate and neutralization value.
[0096] Optionally, according to the test results of the initial information resource recommendation model, the model evaluation index corresponding to the initial information resource recommendation model is determined, and the model evaluation index is used to indicate the performance of the initial information resource recommendation model under various circumstances. Among them, the model evaluation index includes: accuracy, recall and neutralization value, the accuracy is used to indicate the proportion of information resource files that meet the current interest preferences of the target user in the information resource recommendation results generated by the initial information resource recommendation model; the recall rate is used to indicate the proportion of information resource files that meet the current interest preferences of the target user in the information resource recommendation results generated by the initial information resource recommendation model among all the information resource files that actually meet the current interest preferences of the target user; the neutralization value can be represented by the F1 value, and the F1 value is used to neutralize the precision and recall rate.
[0097] Optionally, the model evaluation index of the initial information resource recommendation model is used to evaluate the classification effect of the initial information resource recommendation model.
[0098] S304. Adjust the model parameters of the initial information resource recommendation model according to the model evaluation index, and determine the user preference characteristics to obtain the information resource recommendation model. The model parameters include: learning rate, batch size, and number of hidden layers.
[0099] Optionally, model parameters of the initial information resource recommendation model are adjusted according to the model evaluation indicators of the initial information resource recommendation model. The model parameters include: learning rate, batch size and number of hidden layers. The learning rate is used to indicate the update step of the information resource recommendation model, and the learning rate is related to the update speed of the weight values of the user preference features in the information resource recommendation model; the batch size is used to indicate the size of the number of training samples used in each iterative training during the training process of the information resource recommendation model, and the batch size is related to the training speed, memory usage and convergence of the information resource recommendation model; the number of hidden layers directly affects the performance stability and training time of the information resource recommendation model.
[0100] Optionally, the initial information resource recommendation model uses machine algorithms such as decision trees, random forests, and logistic regression to identify user preference features, and extracts user preference features based on model evaluation indicators of the initial information resource recommendation model. User preference features are used to indicate user preference information for viewing information resource files. User preference features can be divided into attention preference features and importance preference features. Among them, attention preference features are used to indicate the types of information resource files that target users pay special attention to, such as attendance sheets, pay slips, etc., and important preference features are used to indicate important official document logs that target users need to use on a daily basis, such as work reports, etc.
[0101] Optionally, based on the model evaluation index of the initial information resource recommendation model, multiple representative user preference features are selected from the user preference features, or all user preference features in the initial information resource recommendation model are engineered to obtain the most effective user preference features of the initial information resource recommendation model. It is worth noting that after the most effective user preference features are selected based on the model evaluation index of the initial information resource recommendation model, other irrelevant or redundant user preference features are deleted.
[0102] Optionally, the performance of the information resource recommendation model may be changed by transforming the original user preference features or creating new user preference features, such as creating new time series features by calculating differences.
[0103] Optionally, the use of regularization technology in the information resource recommendation model can effectively prevent the information resource recommendation model from overfitting. For example, L1 and L2 regularization can penalize the complexity of the model by adding regularization terms in the loss function, thereby reducing the difficulty of the information resource recommendation model.
[0104] For a possible implementation, see Figure 4 The operation of step S102 may specifically be:
[0105] S401: The information resource recommendation model determines the similarity between the user preference feature and the content feature corresponding to each information resource file according to the user preference feature and the content feature of each information resource file in the information resource database.
[0106] Optionally, the content characteristics of the information resource file are used to characterize the text information contained in the information resource file, and the content characteristics of the information resource file include: the author of the information resource file, keyword information of the information resource file, the date of the information resource file and other data; the user preference characteristics are used to indicate the user's interest preference tags, and the user's interest preference tags may include search keyword information, date, author and other data.
[0107] Optionally, the information resource recommendation model determines a sequence of information resource files that matches the preferences of the target user based on the similarity between the configuration value corresponding to the user preference tag and the configuration value corresponding to each information resource file in the information resource data.
[0108] S402: Sort by similarity and generate information resource recommendation results.
[0109] Optionally, the similarities between the user preference tags and the content features of the information resource files are sorted from large to small, and the information resource recommendation results are generated according to the sequence of similarities from large to small.
[0110] For a possible implementation, see Figure 5 The operation of step S103 may specifically be:
[0111] S501: Determine whether each information resource file is a valid information resource file according to the real-time data of each information resource file in the information resource recommendation result by the target user.
[0112] Optionally, based on the real-time data of each information resource file included in the information resource recommendation result for the target user, it can be determined which information resource files in the information resource recommendation result meet the interest preferences of the target user, and the valid information resource files are used to indicate the information resource files in the information resource recommendation result that are adapted to the current interest preferences of the target user.
[0113] S502: If yes, then update the user preference features in the information resource recommendation model according to the content features of the valid information resource file.
[0114] Optionally, the corresponding user preference features are determined according to the content features of the valid information resource files, and the user preference features and weight values of the user preference features are updated according to the ranking of the valid information resource files in the information resource recommendation results.
[0115] For a possible implementation, see Figure 6 The operation of step S501 may specifically be:
[0116] S601: If the target user views the current information resource file in the information resource recommendation result, and the viewing time of the current information resource file reaches a preset time threshold, the current information resource file is a valid information resource file.
[0117] Optionally, the current information resource file is used to indicate the information resource file currently clicked or jumped to by the user in the information resource recommendation results. The preset time threshold is a time limit set in advance by the user. When the user browses the information resource file page and reaches the preset time threshold, it is determined that the user is interested in the information resource file.
[0118] Optionally, when the target user views the current information resource file in the information resource recommendation result, and the time the user spends viewing the current information resource file reaches a preset time threshold, it is determined that the target user is interested in the current information resource file, and the current information resource file is a valid information resource file.
[0119] S602: If the target user has not viewed the current information resource file in the information resource recommendation result, and / or the viewing time of the current information resource file does not reach the preset time threshold, the current information resource file is an invalid information resource file, and the current information resource file is deleted from the information resource recommendation result.
[0120] Optionally, an invalid information resource file is used to indicate an information resource file in the information resource recommendation result that the target user is not interested in. When the target user has not viewed the current information resource file in the information resource recommendation result, or the target user has viewed the current information resource file in the information resource recommendation result but the length of time the user has viewed the current information resource file has not reached a preset time threshold, it is determined that the target user is not interested in the current information resource file, and the current information resource file is an invalid information resource file, and the invalid information resource file is deleted from the information resource recommendation result.
[0121] It is worth noting that the historical behavior data of each user is analyzed, and the user category is determined based on the similarity between the historical behavior data of each user, and the same information resources are recommended to target users of the same type. By analyzing the content characteristics of the information resource files viewed by the user in the historical period and the historical behavior data of the user, a feature vector between the content of the information resource file and the user is constructed, and based on the feature vector between the content and the user, the similarity between the content of the information resource file and the user is determined. Through the association rule mining algorithm, frequent item set association rules are mined from a large number of historical behavior data of target users, and the correlation between the content of information resource files is identified based on the association rules.
[0122] For a possible implementation, see Figure 7 , the above-mentioned personalized information resource recommendation method further includes:
[0123] S701: Filter historical data of a target user based on a preset time period to obtain historical data of the target user within a target period.
[0124] Optionally, a preset time period is set through a time window method. The preset time period may be 1 day, 7 days, 15 days, 30 days, etc. Based on the preset time period, outdated historical data in the historical data of the target user is filtered out to obtain the historical data of the target user within the target period. The target period is used to indicate a recent period of time. The target period may be within the last 1 day, the last 5 days, the last 15 days, etc. This application does not make any specific limitations on this.
[0125] S702: Extract features from historical data of the target user in the target period to obtain user preference features in the target period.
[0126] Optionally, feature extraction is performed on historical data of the target user within the target period to obtain user preference features within the target period.
[0127] S703: Update the information resource recommendation model in real time according to the user preference characteristics within the target period.
[0128] Optionally, the weight values of the user preference features are updated according to the user preference features in the target period, thereby updating the information resource recommendation model. Updating the information resource recommendation model based only on recent data can increase its sensitivity to user "interest changes" and improve prediction accuracy.
[0129] For a possible implementation, see Figure 8 , the above-mentioned personalized information resource recommendation method further includes:
[0130] S801: Update the weight value of the user preference feature according to a preset forgetting function and the historical data of the target user.
[0131] Optionally, the preset forgetting function can be implemented by the Ebbinghaus forgetting curve, which is used to decay the weight value of the user preference feature corresponding to the historical data of the target user, so as to reduce the influence of the user's historical behavior on the current information resource recommendation results, and increase the weight proportion of the target user's new interests at the current moment. It is worth noting that the preset forgetting function adopts an exponential decay function, and the influence of the user's historical behavior data will decay exponentially over time according to its decay rate, highlighting the influence of recent user behavior data on the update of the information resource model, and weakening the influence of long-term historical data on the information resource model.
[0132] S802: Update the information resource recommendation model in real time according to the updated weight values of the user preference features.
[0133] Optionally, the information resource recommendation model is updated according to the updated weight value of the user preference feature.
[0134] Fig. 9A complete flowchart of the personalized information resource recommendation method provided for this application is available at Fig. 9 The specific process of the personalized information resource recommendation method provided in the embodiment of the present application is: S1, collecting user historical data; S2, building an information resource recommendation model based on the user's historical data; S3, generating information resource recommendation results through the information resource recommendation model; S4, judging whether the user has viewed the recommended content; S5, if not, re-collecting the user's historical data; S6, if yes, further judging whether the user's viewing time of the recommended content reaches a preset time threshold; S7, if no, re-collecting the user's historical data; S8, if yes, scoring the recommended content through the user feedback mechanism.
[0135] Fig.10 A flowchart of a method for updating configuration item information provided by this application, see Fig.10 , the specific process of updating virtual machine configuration item information provided by the embodiment of the present application is as follows: S1. After the virtual machine is created, the initial configuration file of the virtual machine is obtained, and the initial configuration file includes: a virtual machine configuration address, a virtual machine communication protocol and multiple initial configuration items; S2. The running agent component sends a virtual machine registration request to the external cluster according to the virtual machine configuration address in the initial configuration file; S3. After the external cluster receives the virtual machine registration request, the running agent component queries the current configuration file of the virtual machine from the external cluster; S4. The running agent component determines whether there is a target configuration item to be updated for the virtual machine according to the current configuration file and the initial configuration file; S5. If yes, the running agent component updates the configuration value corresponding to the target configuration item according to the type of the target configuration item to be updated; S51. If the target configuration item to be updated is a configuration item from the inside of the virtual machine, the target configuration item to be updated is sent to the external cluster, and the configuration value of the target configuration item in the external cluster is synchronously updated; S52. If the target configuration item to be updated is a configuration item from the external cluster, the configuration value of the initial configuration item corresponding to the target configuration item in the initial configuration file of the virtual machine is updated according to the configuration value corresponding to the target configuration item to be updated.
[0136] Fig.11 A flowchart of a method for determining a target configuration item to be updated provided by this application, see Fig.11A specific process for determining a target configuration item to be updated provided by an embodiment of the present application is as follows: S1101. Determine a configuration value corresponding to each current configuration item according to a label of each current configuration item in the current configuration file; determine a configuration value corresponding to each initial configuration item according to a label of each initial configuration item in the initial configuration file; compare the configuration value corresponding to each current configuration item with the configuration value of the initial configuration item corresponding to the current configuration item according to the label of each current configuration item and the label of each initial configuration item; S1102. Determine the target configuration item to be updated of the virtual machine according to the comparison result of the configuration value corresponding to the current configuration item and the configuration value of the initial configuration item.
[0137] The following describes the apparatus, device, and computer-readable storage medium used to execute the personalized information resource recommendation method provided in this application. The specific implementation process and technical effects are described above and will not be repeated below.
[0138] Fig.12 This is a schematic diagram of the structure of a personalized information resource recommendation device provided in an embodiment of the present application. Fig.12 , the device comprises:
[0139] The acquisition module 1201 is used to acquire the historical data of the target user, the historical data including: historical behavior data, historical social data, user portrait data, historical comment data and historical feedback data;
[0140] A construction module 1202 is used to construct an information resource recommendation model based on the historical data of the target user, and generate an information resource recommendation result by the information resource recommendation model, wherein the information resource recommendation result includes at least one information resource file;
[0141] The updating module 1203 is used to obtain the real-time data of each information resource file in the information resource recommendation result of the target user, and update the information resource recommendation model according to the real-time data to obtain the target information resource recommendation model, the real-time data includes: current behavior data, current social data, current comment data and current feedback data;
[0142] The generating module 1204 is used to generate and output the target information resource recommendation result from the target information resource model.
[0143] In an optional implementation manner, the construction module 1202 is specifically used for:
[0144] Perform data cleaning and data preprocessing on the historical data of target users to obtain valid historical data;
[0145] Build an information resource recommendation model based on valid historical data.
[0146] In an optional implementation manner, the construction module 1202 is further specifically used for:
[0147] Divide the valid historical data into training data set and test data set;
[0148] An initial information resource recommendation model is obtained through training and learning based on the training data set, and the initial information resource recommendation model is tested through the test data set;
[0149] According to the test results of the initial information resource recommendation model, determine the model evaluation indicators corresponding to the initial information resource recommendation model, and the model evaluation indicators include: accuracy, recall rate and neutralization value;
[0150] The model parameters of the initial information resource recommendation model are adjusted according to the model evaluation indicators, and the user preference characteristics are determined to obtain the information resource recommendation model. The model parameters include: learning rate, batch size and number of hidden layers.
[0151] In an optional implementation manner, the construction module 1202 is further specifically used for:
[0152] The information resource recommendation model determines the similarity between the user preference feature and the content feature corresponding to each information resource file according to the user preference feature and the content feature of each information resource file in the information resource database;
[0153] According to the order of similarity, information resource recommendation results are generated.
[0154] In an optional implementation manner, the updating module 1203 is specifically used for:
[0155] Determine whether each information resource file is a valid information resource file according to the real-time data of each information resource file in the information resource recommendation result by the target user;
[0156] If so, the user preference features in the information resource recommendation model are updated according to the content features of the valid information resource files.
[0157] In an optional implementation manner, the updating module 1203 is further configured to:
[0158] If the target user views the current information resource file in the information resource recommendation result, and the viewing time of the current information resource file reaches the preset time threshold, the current information resource file is a valid information resource file;
[0159] If the target user has not viewed the current information resource file in the information resource recommendation results, and / or the viewing time of the current information resource file does not reach the preset time threshold, the current information resource file is an invalid information resource file and is deleted from the information resource recommendation results.
[0160] In an optional implementation manner, the updating module 1203 may also be specifically used for:
[0161] Filter the historical data of the target user based on a preset time period to obtain the historical data of the target user within the target period;
[0162] Extract features from the historical data of the target user in the target period to obtain the user preference features in the target period;
[0163] The information resource recommendation model is updated in real time based on user preference characteristics within the target period.
[0164] In an optional implementation manner, the updating module 1203 may also be specifically used for:
[0165] Update the weight value of the user preference feature according to the preset forgetting function and the historical data of the target user;
[0166] The information resource recommendation model is updated in real time according to the updated weight values of user preference features.
[0167] The above-mentioned device is used to execute the method provided by the aforementioned embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.
[0168] The above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASICs), or one or more microprocessors, or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).
[0169] Fig.13 is a schematic diagram of the structure of a computer device provided in an embodiment of the present application. Fig.13 The computer device includes: a memory 1301 and a processor 1302. The memory 1301 stores a computer program that can be run on the processor 1302. When the processor 1302 executes the computer program, the steps in any of the above method embodiments are implemented.
[0170] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.
[0171] Optionally, the present application also provides a program product, such as a computer-readable storage medium, including a program, which is used to execute any of the above-mentioned personalized information resource recommendation method embodiments when executed by a processor.
[0172] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0173] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0174] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.
[0175] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to perform some steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (English: Read-Only Memory, abbreviated: ROM), random access memory (English: Random Access Memory, abbreviated: RAM), disk or optical disk and other media that can store program codes.
[0176] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
[0177] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. A personalized information resource recommendation method, characterized in that: The method comprises: Obtain historical data of the target user, the historical data including: historical behavior data, historical social data, user portrait data, historical comment data, and historical feedback data; Building an information resource recommendation model based on historical data of the target user, and generating an information resource recommendation result by the information resource recommendation model, wherein the information resource recommendation result includes at least one information resource file; Acquire real-time data of each information resource file in the information resource recommendation result of the target user, and update the information resource recommendation model according to the real-time data to obtain a target information resource recommendation model, wherein the real-time data includes: current behavior data, current social data, current comment data and current feedback data; The target information resource model generates and outputs a target information resource recommendation result.
2. The personalized information resource recommendation method according to claim 1, characterized in that: The step of constructing an information resource recommendation model based on the historical data of the target user includes: Perform data cleaning and data preprocessing on the historical data of target users to obtain valid historical data; An information resource recommendation model is constructed based on the valid historical data.
3. The personalized information resource recommendation method according to claim 2, characterized in that: The step of constructing an information resource recommendation model based on the valid historical data includes: Dividing the valid historical data into a training data set and a test data set; An initial information resource recommendation model is obtained through training and learning based on the training data set, and the initial information resource recommendation model is tested through the test data set; Determine, according to the test results of the initial information resource recommendation model, a model evaluation index corresponding to the initial information resource recommendation model, wherein the model evaluation index includes: accuracy, recall rate and neutralization value; The model parameters of the initial information resource recommendation model are adjusted according to the model evaluation index, and the user preference characteristics are determined to obtain the information resource recommendation model, wherein the model parameters include: learning rate, batch size, and number of hidden layers.
4. The personalized information resource recommendation method according to claim 1, characterized in that: The generating of information resource recommendation results by the information resource recommendation model includes: The information resource recommendation model determines the similarity between the user preference feature and the content feature corresponding to each information resource file according to the user preference feature and the content feature of each information resource file in the information resource database; The information resource recommendation results are generated based on the order of the similarities.
5. The personalized information resource recommendation method according to claim 1, characterized in that: The updating of the information resource recommendation model according to the real-time data to obtain a target information resource recommendation model includes: Determining whether each of the information resource files is a valid information resource file according to the real-time data of each information resource file in the information resource recommendation result by the target user; If so, the user preference features in the information resource recommendation model are updated according to the content features of the valid information resource files.
6. The personalized information resource recommendation method according to claim 5, characterized in that: The step of determining whether each of the information resource files is a valid information resource file according to the real-time data of each information resource file in the information resource recommendation result by the target user comprises: If the target user views the current information resource file in the information resource recommendation result, and the viewing time of the current information resource file reaches the preset time threshold, the current information resource file is a valid information resource file; If the target user has not viewed the current information resource file in the information resource recommendation result, and / or the viewing time of the current information resource file does not reach the preset time threshold, the current information resource file is an invalid information resource file and is deleted from the information resource recommendation result.
7. The personalized information resource recommendation method according to claim 1, characterized in that: Also includes: Filter the historical data of the target user based on a preset time period to obtain the historical data of the target user within the target period; Extract features from the historical data of the target user in the target period to obtain the user preference features in the target period; The information resource recommendation model is updated in real time according to the user preference characteristics within the target period.
8. The personalized information resource recommendation method according to claim 1, characterized in that: Also includes: Update the weight value of the user preference feature according to the preset forgetting function and the historical data of the target user; The information resource recommendation model is updated in real time according to the updated weight values of the user preference features.
9. A personalized information resource recommendation device, characterized in that: The device comprises: An acquisition module is used to acquire historical data of a target user, wherein the historical data includes historical behavior data, historical social data, user portrait data, historical comment data, and historical feedback data; A construction module, used to construct an information resource recommendation model according to the historical data of the target user, and generate an information resource recommendation result by the information resource recommendation model, wherein the information resource recommendation result includes at least one information resource file; An updating module, used to obtain real-time data of each information resource file in the information resource recommendation result of the target user, and update the information resource recommendation model according to the real-time data to obtain a target information resource recommendation model, wherein the real-time data includes: current behavior data, current social data, current comment data and current feedback data; The generation module is used to generate and output the target information resource recommendation result based on the target information resource model.
10. A computer device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method described in any one of claims 1 to 8 are implemented.