A method for intelligent recommendation of teaching courseware

Abstract

The present invention provides a method for intelligent recommendation of teaching courseware, which relates to the field of data processing technology, determines historical recommendation data for teaching courseware recommendation, mines a baseline recommendation system, and based on the baseline recommendation system, performs cross-platform federated integrated learning, supervises the training of intelligent recommendation components, interacts with personalized learning features and learning chains of target users, combines the intelligent recommendation components, initializes the baseline recommendation system, performs platform retrieval, determines the courseware recommendation sequence and displays it on a terminal interface. The method is used to solve the technical problems existing in the prior art of lacking a comprehensive, multi-dimensional comprehensive recommendation framework, being unable to fully mine users' deep-seated needs, eliminating technical barriers to cross-platform recommendations, and being unable to achieve comprehensive and accurate personalized recommendations. The method can effectively improve the accuracy of teaching courseware recommendations and the degree of fit with user needs.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for intelligently recommending teaching courseware. Background Art

[0002] With the digital transformation of education, the management and recommendation of teaching courseware has become increasingly important. Existing teaching courseware recommendation methods usually rely on predicting the courseware that users may be interested in in the future by analyzing their past learning records based on their historical behavior.

[0003] Although existing recommendation technologies can improve the relevance of recommendations to a certain extent, they have certain limitations. For example, they are easily affected by the sparsity of recommendation samples and are difficult to provide personalized and efficient recommendation results. In addition, existing technologies also face challenges in multi-platform and cross-system integrated applications, making it difficult to achieve more accurate recommendation decisions.

[0004] The existing recommendation system lacks a comprehensive, multi-dimensional integrated recommendation framework, is unable to fully tap into users' deep-seated needs, eliminate the technical barriers to cross-platform recommendations, and cannot achieve comprehensive and accurate personalized recommendations. Summary of the invention

[0005] The present application provides a method for intelligently recommending teaching courseware, which is used to solve the technical problems existing in the prior art, such as the lack of a comprehensive, multi-dimensional integrated recommendation framework, the inability to fully tap the deep-seated needs of users, the elimination of technical barriers to cross-platform recommendations, and the inability to achieve comprehensive and accurate personalized recommendations.

[0006] In view of the above problems, the present application provides a method for intelligent recommendation of teaching courseware, including: determining historical recommendation data for teaching courseware recommendation, mining a baseline recommendation system, wherein the baseline recommendation system uses mining decomposition of attribute matrices and time series compensation as mining methods to determine a general concept system; based on the baseline recommendation system, supervised training of an intelligent recommendation component is performed through cross-platform federated integrated learning, wherein the training phase includes a first initialization phase and a second recommendation phase; interacting with the personalized learning characteristics and learning chain of the target user, combining the intelligent recommendation component, initializing the baseline recommendation system, performing platform retrieval, and determining a courseware recommendation sequence; and displaying the courseware recommendation sequence on a terminal interface.

[0007] Furthermore, mining the baseline recommendation system includes: calling historical recommendation data to determine the first recommendation sample by collaborative filtering based on learning type, and determining the second recommendation sample by collaborative filtering based on courseware type, wherein user courseware under similar learning type is used as filtering criteria, and user historical courseware characteristics are used as filtering criteria; mining the recommendation attributes of the first recommendation sample to determine the first recommendation matrix, and mining the recommendation attributes of the second recommendation sample to determine the second recommendation matrix; and fusing the first recommendation matrix with the second recommendation matrix to determine the attribute matrix.

[0008] Furthermore, after determining the attribute matrix, it includes: performing correlation mining on the attribute items in the attribute matrix and the mapping samples to determine the implicit meaning of the correlation; decomposing the attribute matrix based on the implicit meaning of the correlation to determine the first decomposition matrix and the second decomposition matrix, wherein the correlation between the matrix items and the courseware recommendation is used as the first decomposition standard, and the weight of the matrix items in the attribute matrix is ​​used as the second decomposition standard; for multiple teaching stages in the entire teaching cycle, performing stage correlation feature mining to determine the time series relationship, wherein the stage correlation feature is the attribute influence of the previous stage on the next stage; based on the first decomposition matrix, the second decomposition matrix and the time series relationship, constructing the baseline recommendation system.

[0009] Furthermore, supervised training of an intelligent recommendation component includes: embedding the baseline recommendation system as an intelligent recommendation module, wherein the intelligent recommendation module includes a system initialization block and a recommendation decision block; decentralizing the intelligent recommendation module to each platform, calling the platform's internal recommendation data stream for training, and determining N intelligent recommendation modules; integrating the N intelligent recommendation modules to construct the intelligent recommendation component.

[0010] Furthermore, the personalized learning features and learning chain of the interactive target user include: determining the courseware retrieval information of the target user, performing semantic analysis, and determining the personalized learning features; performing cross-platform interactive calls, determining the historical learning records of the target user, and building a learning chain.

[0011] Furthermore, determining the courseware recommendation sequence includes: through the intelligent recommendation component of the first platform, based on the personalized learning characteristics and learning chain, making parallel decisions in the N intelligent recommendation modules to determine N recommendation sequences, wherein the first platform is a retrieval platform for target users; traversing the N recommendation sequences, performing comprehensive sorting and screening by a preset number to determine the courseware recommendation sequence.

[0012] Furthermore, parallel decision making is performed in the N intelligent recommendation modules, including: assisting the first intelligent recommendation module, extracting attribute features in the system initialization block for the personalized learning features and learning chain, and determining a first personalized recommendation system, wherein the first intelligent recommendation module is any one of the N intelligent recommendation modules; transferring the first personalized recommendation system to the recommendation decision block, taking the courseware pool of the first platform as the recommendation target, and determining a first recommendation sequence.

[0013] Furthermore, taking the courseware pool of the first platform as the recommendation target, a first recommendation sequence is determined, including: the courseware pool includes the teaching courseware resources of the first platform, and the teaching courseware resources of the platform that has established resource interaction and sharing with the first platform; taking the courseware pool as the recommendation target, a recommendation decision is made to determine the preset number of pre-recommended courseware; the pre-recommended courseware is positively serialized and integrated based on priority to determine the first recommendation sequence, wherein the pre-recommended courseware on the platform that has established resource interaction and sharing with the first platform is displayed as a cross-platform link.

[0014] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0015] The embodiment of the present application provides a method for intelligent recommendation of teaching courseware, which determines the historical recommendation data of teaching courseware recommendation, mines the baseline recommendation system, and based on the baseline recommendation system, supervises the training of the intelligent recommendation component by performing cross-platform federated integrated learning, and the training stage includes a first initialization stage and a second recommendation stage; the personalized learning characteristics and learning chain of the interactive target user are combined with the intelligent recommendation component to initialize the baseline recommendation system, perform platform retrieval, and determine the courseware recommendation sequence; and display the courseware recommendation sequence on the terminal interface, which is used to solve the technical problems existing in the prior art of lacking a comprehensive, multi-dimensional comprehensive recommendation framework, being unable to fully mine the deep-seated needs of users, eliminating the technical barriers of cross-platform recommendation, and being unable to achieve comprehensive and accurate personalized recommendation, and can effectively improve the accuracy of teaching courseware recommendation and the degree of fit with user needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flow chart of a method for intelligently recommending teaching courseware is provided for this application;

[0017] Figure 2 A schematic diagram of the construction process of an intelligent recommendation component in an intelligent recommendation method for teaching courseware is provided for this application;

[0018] Figure 3 A schematic diagram of the process of obtaining the first recommendation sequence in a method for intelligently recommending teaching courseware is provided for this application. DETAILED DESCRIPTION

[0019] The present application provides a method for intelligent recommendation of teaching courseware, determines the historical recommendation data for teaching courseware recommendation, mines the baseline recommendation system, and based on the baseline recommendation system, performs cross-platform federated integrated learning, supervises the training of intelligent recommendation components, interacts with the personalized learning characteristics and learning chain of target users, combines the intelligent recommendation components, initializes the baseline recommendation system, performs platform retrieval, determines the courseware recommendation sequence and displays it on the terminal interface. The application is used to solve the technical problems existing in the prior art, such as the lack of a comprehensive, multi-dimensional comprehensive recommendation framework, the inability to fully explore the deep-seated needs of users, the elimination of technical barriers to cross-platform recommendations, and the inability to achieve comprehensive and accurate personalized recommendations.

[0020] Example: Figure 1 As shown, the present application provides a method for intelligently recommending teaching courseware, the method comprising:

[0021] S1: Determine the historical recommendation data for teaching courseware recommendation and mine the baseline recommendation system, wherein the baseline recommendation system uses the mining decomposition of attribute matrix and time series compensation as the mining method to determine the general concept system;

[0022] In the embodiment of the present application, the historical recommendation data for teaching courseware recommendation refers to the relevant data of teaching courseware used, browsed or recommended by users in the past, usually including the behavioral records of different users on the teaching content such as clicks, downloads, viewing time, and ratings. The baseline recommendation system refers to a basic recommendation system established by analyzing the historical recommendation data and combining different attribute dimensions (such as learning type, courseware type, etc.), that is, a general concept system, a unified conceptual framework suitable for different scenarios and platforms extracted during data mining, based on user behavior, courseware attributes, time changes, etc., which can provide reasonable recommendation basis in different learning scenarios, and has the ability to be applied across platforms, providing a preliminary framework and basis for subsequent personalized recommendations and adjustments.

[0023] Among them, the mining and decomposition of the attribute matrix refers to converting various attributes of the courseware (such as course type, knowledge points, difficulty, learning time, etc.) into a matrix form, and extracting the potential information hidden behind the data through matrix decomposition to reveal the potential correlation between the courseware and user behavior. Time series compensation considers changes in the time dimension. For example, some courseware may be more attractive to users in a specific time period, or a certain stage in the user's learning process requires a specific type of courseware. By introducing the time factor, it ensures that the recommendation results can match the user's learning cycle and learning progress.

[0024] Specifically, for the historical recommendation data, by mapping the different attributes of the courseware into a matrix form, the intrinsic relationship between each courseware and different user behaviors is captured, and by refining the most basic attribute concepts, the baseline recommendation system is formed. For example, if a user prefers to choose a certain type of mathematics courseware in the past, the matrix decomposition algorithm can help discover the user's potential interest in mathematics courseware and recommend similar courseware to him. At the same time, the introduction of time series compensation optimizes the recommendation effect. Taking into account the user's learning process and time factors, some courseware may become more relevant at different stages of the user's learning cycle. For example, after a student completes a basic mathematics course, he may become interested in advanced mathematics courseware. Through time series compensation, the recommendation sequence can be dynamically adjusted to match the student's learning progress.

[0025] By performing attribute association mining, matrix decomposition and time adjustment on data, not only the accuracy of recommendation decisions is improved, but also the recommendation process has strong adaptability in different learning environments and platforms.

[0026] Among them, mining the baseline recommendation system, step S1 of this application includes:

[0027] Call historical recommendation data to determine the first recommendation sample through collaborative filtering based on learning type, and determine the second recommendation sample through collaborative filtering based on courseware type, wherein user courseware under similar learning type is used as filtering criteria, and user historical courseware characteristics are used as filtering criteria; perform recommendation attribute mining on the first recommendation sample to determine the first recommendation matrix, and perform recommendation attribute mining on the second recommendation sample to determine the second recommendation matrix; fuse the first recommendation matrix and the second recommendation matrix to determine the attribute matrix.

[0028] In an embodiment of the present application, collaborative filtering based on learning type is performed, using user courseware under similar learning types as filtering criteria, analyzing the behavior of similar users under the same learning type, and inferring the courseware that the current user may be interested in. Exemplarily, the learning type can be a course category (such as mathematics, physics, chemistry, etc.) or a learning goal (such as basic learning, advanced learning, etc.). By collaborative filtering based on courseware type, using the user's historical courseware characteristics as filtering criteria, analyzing the user's interactive behavior on different courseware types (such as text, video, image, etc.) to make recommendations. By understanding the user's preference for a certain type of courseware, the same type of courseware can be better recommended.

[0029] Triggered by the learning type and courseware type, the historical recommendation data is divided into the first recommendation sample and the second recommendation sample. For example, if a user has mainly studied mathematics-related courses in the past, the system will find other users with similar learning trajectories and recommend related courses based on their choices, that is, user courses under similar learning types. Similarly, collaborative filtering based on courseware type focuses on analyzing users' preferences for different courseware types (such as videos, texts, charts, etc.). For example, if a user likes highly interactive video courses, the system will give priority to recommending other video courses as the second recommendation sample.

[0030] The recommended attribute mining refers to extracting useful feature information by analyzing the attribute data of the recommended samples, thereby forming the attribute matrix.

[0031] Specifically, the first and second recommended samples are mined for recommended attributes to extract features related to user needs. For the first recommended sample, the system analyzes attributes related to the learning type (such as course difficulty, courseware duration, etc.) to form a first recommendation matrix; for the second recommended sample, the system analyzes attributes related to the courseware type (such as courseware format, content coverage, etc.) to form a second recommendation matrix.

[0032] By integrating the first recommendation matrix and the second recommendation matrix, the final attribute matrix is ​​constructed. This matrix will combine the recommendation information of learning type and courseware type to generate more personalized and accurate recommendation results. By integrating multi-dimensional data, the attribute matrix can comprehensively consider the user's multiple preferences, ensuring that the recommendation system can accurately identify user needs and provide the most relevant courseware recommendations.

[0033] After determining the attribute matrix, step S1 of the present application includes:

[0034] Perform correlation mining on the attribute items and mapping samples in the attribute matrix to determine the implicit meaning of the correlation; based on the implicit meaning of the correlation, decompose the attribute matrix to determine the first decomposition matrix and the second decomposition matrix, wherein the correlation between the matrix items and the courseware recommendation is used as the first decomposition standard, and the weight of the matrix items in the attribute matrix is ​​used as the second decomposition standard; for multiple teaching stages in the entire teaching cycle, perform stage correlation feature mining to determine the time series relationship, wherein the stage correlation feature is the attribute influence of the previous stage on the next stage; based on the first decomposition matrix, the second decomposition matrix and the time series relationship, construct the baseline recommendation system.

[0035] First, the correlation between the attribute items in the attribute matrix and the mapping samples is mined to deeply analyze the implicit relationship between the courseware attributes and user preferences. The mapping samples are sample data obtained by combining historical recommendation data with the attribute matrix, that is, extracted from multiple recommendation sources and represented in the form of a matrix to help the system identify and evaluate the recommendation attributes of different courseware.

[0036] By mining the association between the attribute matrix and the mapping samples, that is, conducting data mining between the attribute matrix and the mapping samples, the potential connections and interactions between different attribute items can be identified, such as the potential connection between courseware attributes (such as difficulty, duration, type, etc.) and user learning behavior. For example, if a user shows a high interest in learning certain math courses, the system may find that the difficulty of these courses matches the user's learning ability, thereby improving the accuracy of the recommendation.

[0037] The implicit meaning of association refers to the features or relationships revealed by association mining that are not directly apparent but have an important impact on the recommendation system. For example, the difficulty of a course may be related to the user's learning ability and learning history, but these associations may not be obvious and can only be revealed through data mining.

[0038] The attribute matrix is ​​decomposed based on the implicit meaning of the association. Specifically, the attribute matrix is ​​decomposed into two sub-matrices, namely the first decomposition matrix and the second decomposition matrix, through matrix decomposition technology. The first decomposition matrix mainly extracts the relationship between user preferences and courseware attributes by measuring the correlation between matrix items and courseware recommendations; the second decomposition matrix focuses on analyzing the weights of matrix items in the attribute matrix to help the system identify which courseware attributes have a greater impact on user choices. Through matrix decomposition, the system can effectively extract and reconstruct users' potential interests and provide accurate data support for personalized recommendations.

[0039] On this basis, the entire teaching cycle is divided into multiple stages, and stage-related feature mining is carried out to determine the correlation between different teaching stages in the entire teaching cycle, so as to measure the attribute influence of the previous stage on the next stage as the time series relationship, wherein the stage-related feature refers to the influence of the learning content of the previous stage on the learning content of the subsequent stage in the teaching process.

[0040] That is, the time series relationship refers to the changing pattern of course recommendations in the time dimension. Since different teaching stages and learning progress have different requirements for recommendations, the impact of time series needs to be considered to ensure the timeliness and relevance of recommendations.

[0041] For example, the content learned in the basic mathematics stage will affect the recommendation of subsequent advanced mathematics courses. The stage-related characteristics reveal the interdependence between different stages of teaching. For example, after completing the basic stage of learning, students may have a stronger demand for advanced courses. At this time, the system needs to adjust the recommended content based on the time series relationship to meet the needs of students at different learning stages. Through this staged adjustment, the system can ensure that the recommended content is highly relevant and timely at different time points.

[0042] Finally, based on the comprehensive analysis of the first decomposition matrix, the second decomposition matrix and the time series relationship, a comprehensive baseline recommendation system is constructed. It combines multiple information such as the user's learning progress, courseware attributes, and phased needs to provide users with more personalized and accurate courseware recommendations. In the entire recommendation process, the combination of the decomposition of the attribute matrix and the time series relationship enables the recommendation system to continuously adjust the recommendation strategy according to the dynamic needs of users, optimize the recommendation effect, and ensure the relevance and timeliness of the recommended content.

[0043] S2: Based on the baseline recommendation system, supervise the training of intelligent recommendation components by performing cross-platform federated ensemble learning, wherein the training phase includes a first initialization phase and a second recommendation phase;

[0044] In the present invention, based on the constructed baseline recommendation system, a cross-platform federated integrated learning method is used, that is, data sharing and collaborative learning is performed between multiple platforms, the local data of each platform is used as a training sample, and training is performed based on the data characteristics of different platforms. Multiple intelligent recommendation modules are trained, and the learning results are fed back to the central server for model integration. By integrating models from different platforms, the advantages of each platform can be utilized to comprehensively obtain a more accurate recommendation model.

[0045] The training process specifically includes: the first initialization stage is mainly to complete the initial configuration of the model. Since the baseline recommendation system is a universal attribute concept system, the baseline recommendation system is initialized according to the recommendation requirements. On this basis, the second recommendation stage training is performed, that is, after the first initialization stage is completed, sample-driven training is performed with the courseware recommendation decision as the goal to build the intelligent recommendation component.

[0046] Through these two stages of training, the intelligent recommendation component has been continuously improved. It can accurately analyze user behavior and generate high-quality recommendation sequences. It has the ability to automatically analyze user behavior, learn user preferences, and generate personalized recommendations. The component can infer the courseware content that meets the user's needs based on the characteristics of the user input (such as learning needs, historical behavior, etc.).

[0047] Among them, Figure 2As shown, supervised training of intelligent recommendation components, step S2 of this application includes:

[0048] The baseline recommendation system is built-in as an intelligent recommendation module, and the intelligent recommendation module includes a system initialization block and a recommendation decision block; the intelligent recommendation module is decentralized to each platform, and the internal recommendation data stream of the platform is called for training to determine N intelligent recommendation modules; the N intelligent recommendation modules are integrated to construct the intelligent recommendation component.

[0049] First, the baseline recommendation system is built-in as the core framework of the intelligent recommendation module. The baseline recommendation system provides a preliminary recommendation structure. The system initialization block in the intelligent recommendation module configures the baseline recommendation system to meet the personalized status of the current user and completes the preliminary settings of the system, including data input and output paths, initial assignment of attribute parameters, etc.

[0050] Next, the intelligent recommendation module will be decentralized to each platform and trained using the historical recommendation data stream within the platform. The purpose is to allow each platform's intelligent recommendation module to conduct preliminary learning on local data, so that it can adjust and optimize the recommendation model according to the platform's unique data characteristics, that is, generate multiple personalized intelligent recommendation modules based on the data characteristics of different platforms. Each module will be trained and optimized for different user behavior patterns on its own platform to adapt to the specific needs of the platform. Where N is the number of platforms, and the intelligent recommendation module corresponds to the platform one by one.

[0051] By further integrating the N intelligent recommendation modules and combining the intelligent recommendation modules trained on multiple platforms, the system can integrate the advantages of each platform to form a unified cross-platform recommendation system. The integrated intelligent recommendation component can provide more accurate and comprehensive courseware recommendations by integrating the recommendation models of different platforms, and has powerful personalized recommendation capabilities. Through cross-platform data flows and multi-module collaboration, the intelligent recommendation component can provide users with more accurate, flexible and highly personalized courseware recommendations to meet the diverse needs of users on different platforms.

[0052] S3: interact with the personalized learning characteristics and learning chain of the target user, combine the intelligent recommendation component, initialize the baseline recommendation system, perform platform search, and determine the courseware recommendation sequence;

[0053] S4: Displaying the courseware recommendation sequence on a terminal interface.

[0054] In the embodiment of the present application, the personalized learning characteristics and learning chain of the interactive target user are used to obtain and analyze the user's behavior data during the learning process, and identify the user's personalized learning characteristics, such as the input learning recommendation requirements, learning speed, preferred course type, learning ability, etc. At the same time, the user's learning chain is constructed, that is, the user's entire learning activity trajectory from the beginning of learning to the current state.

[0055] Furthermore, the baseline recommendation system provides a framework to initialize the baseline recommendation system through these personalized learning features and learning chains to ensure that the recommendation model can adapt to the learning habits and needs of each user. Further, based on the recommendation decision block in the intelligent recommendation component, platform resource retrieval and recommendation decision are performed, that is, the courses most relevant to the user's learning goals are found in the platform, and they are sorted for recommendation, and the courseware recommendation sequence is determined.

[0056] Preferably, the courseware recommendation sequence is optimized and sorted according to the user's historical behavior, learning progress and current learning needs to ensure that the recommended courseware is most beneficial to the user. At the same time, the generation process of the recommendation sequence takes into account multiple factors, including the user's interests, learning progress and the contribution of each courseware to his learning goals. Finally, the courseware recommendation sequence is displayed on the user terminal interface.

[0057] Among them, the personalized learning characteristics and learning chain of the interactive target user, step S3 of this application includes:

[0058] Determine the courseware retrieval information of the target user, perform semantic analysis, and determine the personalized learning features; perform cross-platform interactive calls, determine the historical learning records of the target user, and build a learning chain.

[0059] In the embodiment of the present application, by determining the courseware retrieval information of the target user, it is usually the query request or task requirement actively input by the user during the learning process. By capturing this information, the user's current learning goals and needs are identified. Further semantic parsing is performed, that is, the text or query content input by the user is analyzed to extract the learning intention and actual needs behind it. Through semantic parsing, keywords, intentions, contextual relationships, etc. can be extracted from complex natural languages, so that the system can accurately understand the user's real needs as the personalized learning feature.

[0060] Next, through the cross-platform interactive calling technology, the historical learning records of the target user on different learning platforms are collected. By obtaining the learning track of the target user on different platforms, these historical learning records include not only the user's behavior on a certain platform, but also the user's learning activities on other platforms, helping the system to fully understand the user's learning status and progress.

[0061] By integrating information, a personalized learning chain is built for the target user, that is, a behavioral sequence that reflects the user's learning path. The learning chain shows the user's learning activities and progress between multiple learning stages. Through this data, the system can determine the user's current learning status and possible future learning needs. This enables recommendation decisions to better track the user's learning progress.

[0062] Wherein, the step S3 of determining the courseware recommendation sequence in this application includes:

[0063] Through the intelligent recommendation component of the first platform, based on the personalized learning features and learning chain, parallel decisions are made in the N intelligent recommendation modules to determine N recommendation sequences, wherein the first platform is a retrieval platform for target users; the N recommendation sequences are traversed, comprehensive sorting and preset number screening are performed to determine the courseware recommendation sequence.

[0064] Specifically, the constructed intelligent recommendation component is decentralized to each platform, and decision processing is performed when there is a recommendation demand on the platform. Specifically, the intelligent recommendation component of the first platform is used to make parallel recommendation decisions in the N intelligent recommendation modules within the intelligent recommendation component through the user's personalized learning characteristics and learning chain, that is, decisions are made based on the user's personalized learning characteristics (such as learning style, interest, learning progress, etc.) and learning chain (i.e., the user's learning history and trajectory, etc.).

[0065] In the parallel decision-making process, the N intelligent recommendation modules perform recommendation decision processing relatively independently, output the courseware recommendation results respectively, and determine N recommendation sequences. The N recommendation sequences are further comprehensively sorted and screened by a preset number. The comprehensive sorting will conduct a comprehensive evaluation of the recommended items of multiple recommendation sequences according to factors such as user preferences and learning progress to determine the comprehensive order, that is, arrange them in order of priority. The preset number is to limit the number of recommendations finally displayed to the user, and can be customized to screen out the top N most suitable courseware or resources from the recommendation sequence as the courseware recommendation sequence.

[0066] Through multi-module concurrent decision-making, the recommendation characteristics of multiple platforms can be taken into account to make decisions, improving the completeness and effectiveness of recommendation decisions. Based on the module decision output, an overall recommendation evaluation is performed to improve the accuracy and personalization of the recommendation, while also enhancing the user's learning experience, allowing them to obtain more efficient and accurate resource recommendations during the learning process.

[0067] Among them, parallel decision making is performed in the N intelligent recommendation modules, and step S3 of the present application includes:

[0068] Assist the first intelligent recommendation module, extract attribute features in the system initialization block for the personalized learning features and learning chain, and determine the first personalized recommendation system, wherein the first intelligent recommendation module is any one of the N intelligent recommendation modules; transfer the first personalized recommendation system to the recommendation decision block, take the courseware pool of the first platform as the recommendation target, and determine the first recommendation sequence.

[0069] Specifically, the first intelligent recommendation module refers to any one of the N intelligent recommendation modules in the intelligent recommendation component. The auxiliary first intelligent recommendation module extracts attribute features in the system initialization block based on the personalized learning features and learning chain of the user, that is, by analyzing the learning behavior and learning path (i.e., learning chain) of the target user, as well as the personalized features such as the user's preferences, interests, and learning progress, extracts key attribute features related to the recommendation decision, and initializes the baseline recommendation system, that is, based on the specific attribute feature values, identifies the attribute concepts in the baseline recommendation system, and determines the first personalized recommendation system so that it matches the target user.

[0070] The first personalized recommendation system is further transferred to the recommendation decision block, and a courseware recommendation decision is made based on the first personalized recommendation system, and the recommended courseware is determined and sorted in a limited manner. Based on the preset number, the recommended courseware that meets the preset number is screened as the first recommendation sequence. Specifically, in the recommendation decision block, the courseware pool of the first platform is used as the target of the recommended resources. The courseware pool refers to the teaching courseware resource library of all available courses on the first platform. In this solution, the courseware pool includes a retrieval platform, that is, all learning resources on the first platform that can be recommended to users, and teaching courseware resources of a platform that has established resource interaction and sharing with the first platform.

[0071] Finally, the first recommendation sequence is based on the decision result of the first intelligent recommendation module. Similarly, the N intelligent recommendation modules are processed in parallel. The recommendation results generated based on personalized learning characteristics and learning chains include the learning resources most needed by users. This recommendation sequence comprehensively analyzes the learning behavior and needs of users to ensure that users are provided with courseware resources that best suit their current learning stage and needs. Through this process, the recommendation system can accurately provide users with customized learning content, improve learning effects and user experience.

[0072] Among them, Figure 3 As shown, taking the courseware pool of the first platform as the recommendation target, determining the first recommendation sequence, step S3 of this application includes:

[0073] The courseware pool includes teaching courseware resources of the first platform and teaching courseware resources of platforms with which resource interaction and sharing is established with the first platform; taking the courseware pool as the recommendation target, recommendation decisions are made to determine the preset number of pre-recommended courseware; the pre-recommended courseware are positively serialized and integrated based on priority to determine the first recommendation sequence, wherein the pre-recommended courseware on the platform with which resource interaction and sharing is established with the first platform is displayed as a cross-platform link.

[0074] In the embodiment of the present application, the courseware pool is a resource library for courseware recommendation, which includes teaching courseware resources of the first platform and teaching courseware resources of external platforms that have established resource sharing relationships with the first platform. Through such resource integration, the courseware pool can provide a more comprehensive and rich teaching content library, ensuring that the recommendation system can obtain teaching resources that users may be interested in from multiple platforms.

[0075] Next, the courseware pool of the first platform is used as a recommendation target to make a recommendation decision, and pre-recommended courseware that meets user needs is screened out, and the preset number is the number of courseware screening. The pre-recommended courseware is further prioritized and integrated as the first recommendation sequence. Preferably, a priority is assigned to each courseware based on factors such as the degree of match between the courseware and user needs, learning progress, and the courseware is sorted according to these priorities. The sorted courseware will form an ordered recommendation list, which is the first recommendation sequence. This process ensures that the order of courseware finally displayed to the user can meet the user's learning needs and interests to the greatest extent possible.

[0076] Among them, when the recommended courseware comes from a platform that has established resource interaction and sharing with the first platform, the recommendation system will display the resources of the external platform through cross-platform link display and provide jump links. This ensures that even if the courseware comes from different platforms, users can directly access and learn through simple operations, optimizes the cross-platform resource utilization efficiency, and improves the user experience. Through the above process, the recommendation system can provide users with accurate and rich learning resources, and ensure the diversity and accessibility of the recommendation results.

[0077] The intelligent recommendation method for teaching courseware provided by this application has the following technical effects:

[0078] 1. By building a baseline recommendation system based on attribute matrix mining and time series compensation. Establish a general recommendation concept system, through attribute matrix decomposition (including relevance and weight decomposition) and time series analysis, dynamically optimize the recommendation strategy according to the user's learning characteristics and behavior changes, and improve the accuracy of recommendations and user satisfaction.

[0079] 2. Based on the cross-platform federated integrated learning framework, the system simultaneously trains intelligent recommendation components on multiple platforms to ensure the multi-platform applicability and generalization ability of the recommendation algorithm. Through cross-platform learning, it is possible to pool the data advantages of different platforms to train a more comprehensive and accurate recommendation model, further improve the personalized effect of recommendations, and avoid the impact of data bias on a single platform.

[0080] 3. Ensure the diversity and comprehensiveness of resources by integrating the teaching courseware resources of the first platform and the teaching resources shared with other platforms. Through cross-platform resource integration, the system can provide users with richer and more extensive learning resources, avoid the limitations of platform resources, and enhance the diversity of recommended content and the accessibility of resources. When the recommended courseware comes from a cross-platform, the system displays it through cross-platform links to ensure that users can easily access the resources of the external platform.

[0081] In summary, through multi-dimensional data mining and analysis, combined with intelligent recommendation algorithms and cross-platform resource integration, the accuracy, personalization and cross-platform adaptability of the recommendation system have been significantly improved. This ensures that users can obtain the most relevant learning resources based on their learning characteristics and progress, and through the collaborative work of multiple platforms, the diversity of resources and access convenience have been further improved.

[0082] Through the above-mentioned detailed description of a method for intelligently recommending teaching courseware in this specification, those skilled in the art can clearly understand a method for intelligently recommending teaching courseware in this embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0083] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligently recommending teaching courseware, characterized in that: The method comprises: Determine the historical recommendation data for teaching courseware recommendation and mine the baseline recommendation system, wherein the baseline recommendation system uses the mining decomposition of attribute matrix and time series compensation as the mining method to determine the general concept system; Based on the baseline recommendation system, supervise the training of intelligent recommendation components by performing cross-platform federated integrated learning, wherein the training phase includes a first initialization phase and a second recommendation phase; Among them, supervised training of intelligent recommendation components includes: The baseline recommendation system is built-in as an intelligent recommendation module, and the intelligent recommendation module includes a system initialization block and a recommendation decision block; Decentralize the intelligent recommendation module to each platform, call the internal recommendation data stream of the platform for training, and determine N intelligent recommendation modules; Integrate the N intelligent recommendation modules to construct the intelligent recommendation component, wherein the comprehensive ranking of the N recommendation sequences of the N intelligent recommendation modules in the intelligent recommendation component is output; Calling the personalized learning characteristics and learning chain of the target user, combining the intelligent recommendation component, initializing the baseline recommendation system, performing platform search, and determining the courseware recommendation sequence; Displaying the courseware recommendation sequence on a terminal interface; After determining the attribute matrix, the following steps are included: Performing correlation mining on each attribute item in the attribute matrix and a mapping sample to determine the implicit meaning of the correlation, wherein the mapping sample is sample data obtained by combining historical recommendation data with the attribute matrix, extracted from multiple recommendation sources and represented in the form of a matrix; Based on the implicit meaning of the association, the attribute matrix is ​​decomposed to determine a first decomposition matrix and a second decomposition matrix, wherein the relevance between the matrix items and the courseware recommendation is used as the first decomposition standard, and the weight of the matrix items in the attribute matrix is ​​used as the second decomposition standard; For multiple teaching stages in the entire teaching cycle, we conduct stage-related feature mining to determine the time series relationship. The stage-related feature is the attribute influence of the previous stage on the next stage. The baseline recommendation system is constructed based on the first decomposition matrix, the second decomposition matrix and the time series relationship.

2. A teaching courseware intelligent recommendation method as claimed in claim 1, characterized in that: Mining baseline recommendation systems, including: Calling historical recommendation data, determining the first recommendation sample by collaborative filtering based on learning type, and determining the second recommendation sample by collaborative filtering based on courseware type, wherein user courseware under similar learning type is used as filtering criteria, and user historical courseware characteristics are used as filtering criteria; Performing recommendation attribute mining on the first recommended sample to determine a first recommendation matrix, and performing recommendation attribute mining on the second recommended sample to determine a second recommendation matrix; The first recommendation matrix and the second recommendation matrix are integrated to determine the attribute matrix.

3. The method for intelligently recommending teaching courseware according to claim 1, characterized in that: Personalized learning features and learning chains of interactive target users, including: Determine the courseware retrieval information of the target user, perform semantic analysis, and determine the personalized learning features; Conduct cross-platform interactive calls, determine the historical learning records of target users, and build a learning chain.

4. A teaching courseware intelligent recommendation method as claimed in claim 3, characterized in that: The step of determining the courseware recommendation sequence includes: Through the intelligent recommendation component of the first platform, based on the personalized learning features and the learning chain, parallel decision making is performed in the N intelligent recommendation modules to determine N recommendation sequences, wherein the first platform is a search platform for target users; The N recommendation sequences are traversed, and comprehensive sorting and preset number screening are performed to determine the courseware recommendation sequence.

5. A teaching courseware intelligent recommendation method as claimed in claim 4, characterized in that: Parallel decision making is performed in the N intelligent recommendation modules, including: Assisting the first intelligent recommendation module to extract attribute features in the system initialization block based on personalized learning features and learning chains, and determine the first personalized recommendation system; The first personalized recommendation system is transferred to the recommendation decision block, and the courseware pool of the first platform is used as the recommendation target to determine a first recommendation sequence; Among them, the first intelligent recommendation module is any one of the N intelligent recommendation modules.

6. A teaching courseware intelligent recommendation method as claimed in claim 5, characterized in that: Taking the courseware pool of the first platform as the recommendation target, determining a first recommendation sequence includes: The courseware pool includes the teaching courseware resources of the first platform and the teaching courseware resources of the platform that has established resource interaction and sharing with the first platform; Taking the courseware pool as a recommendation target, making a recommendation decision to determine the preset number of pre-recommended courseware; The pre-recommended courseware is positively serialized and integrated based on the priority to determine the first recommendation sequence, wherein the pre-recommended courseware on the platform that has established resource interactive sharing with the first platform is displayed as a cross-platform link.

Images