Information recommendation method, device and system, electronic equipment and readable storage medium

Abstract

The invention provides an information recommendation method and device, and relates to the technical field of information recommendation, in particular to the technical fields of natural language processing, statistics, machine learning, large language models and the like. According to the specific implementation scheme, an association rule set is determined and sent based on a data set when a learning object completes an initial course; obtaining the interestingness of the user on the association rule set; matching the association rule set with rules in a knowledge base, and determining a matching result of the association rule set; updating the information tuple in the knowledge base based on the interestingness, the matching result and the association rule set; and based on the information tuple in the knowledge base, determining a recommended course, and sending the recommended course to the learning object.

Description

Technical Field

[0001] The present disclosure relates to the field of information recommendation technology, particularly to natural language processing, machine learning, and large-scale language models. In particular, it provides an information recommendation method and apparatus, an information recommendation system, an electronic device, and a computer-readable storage medium. Background Art

[0002] Data mining technology is increasingly used in online learning and web-based adaptive education systems. When data mining is applied in online learning, the useful information discovered can be directly used by course students or authors.

[0003] Applying data mining to online learning, particularly object-centric approaches aimed at improving courses, involves a series of obstacles to overcome. Data mining tools typically prioritize power and flexibility over ease of use. Most current data mining tools are too complex for educators, and their features do not cover the range of potential needs. On the one hand, there are many e-learning and web-based adaptive courses to which data mining can be applied, influenced by three key factors: first, the knowledge domain covered by the course; second, the course level (university, secondary, or elementary level, special education, or other types of courses); and finally, the course's difficulty level—whether it is a basic or beginner, intermediate, advanced, or expert course. On the other hand, the wide range of results that can be obtained based on these factors means that finding universal, reproducible patterns that can be applied to any type of course is quite challenging. Furthermore, educational datasets are typically small, compared to databases used in other data mining fields (such as e-commerce applications), which can involve tens of thousands of customers. Therefore, applying data mining using specific filtering parameters can be problematic when discovering association rules in small databases, where the initial information is insufficient to build a model to infer future recommendations. Summary of the Invention

[0004] The present disclosure provides an information recommendation method and apparatus, an information recommendation system, an electronic device, and a computer-readable storage medium.

[0005] According to a first aspect, an information recommendation method is provided, which includes: determining and sending an association rule set based on a data set when a learning object completes an initial course; obtaining a user's interest in the association rule set; matching the association rule set with rules in a knowledge base to determine a matching result of the association rule set; updating an information tuple in the knowledge base based on the interest, the matching result, and the association rule set; determining a recommended course based on the information tuple in the knowledge base, and sending the recommended course to the learning object.

[0006] According to a second aspect, an information recommendation device is provided, which includes: a determination unit, configured to determine and send an association rule set based on a data set when a learning object completes an initial course; an acquisition unit, configured to obtain a user's interest in the association rule set; a matching unit, configured to match the association rule set with rules in a knowledge base, and determine a matching result of the association rule set; an updating unit, configured to update an information tuple in the knowledge base based on the interest, the matching result and the association rule set; a recommendation unit, configured to determine a recommended course based on the information tuple in the knowledge base, and send the recommended course to the learning object.

[0007] According to a third aspect, an information recommendation system is provided, which includes: a client and a server, wherein the client is used to determine and send an association rule set based on a data set when a learning object completes an initial course; obtain a user's interest in the association rule set; match the association rule set with rules in a knowledge base to determine a matching result of the association rule set; update information tuples in the knowledge base based on the interest, the matching result and the association rule set; determine recommended courses based on the information tuples in the knowledge base, and send the recommended courses to the learning object; the server includes: a network application module and a network service module, wherein the network application module is used to provide an interface for managing the knowledge base so as to update the knowledge base through the interface; the network service module is used to provide the updated knowledge base to the client.

[0008] According to a fourth aspect, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described in any implementation manner of the first aspect.

[0009] According to a fifth aspect, a non-transitory computer-readable storage medium storing computer instructions is provided, where the computer instructions are used to cause a computer to execute the method as described in any implementation of the first aspect.

[0010] The information recommendation method and device provided by the embodiments of the present disclosure first determine and send an association rule set based on the data set when the learning object completes the initial course; secondly, obtain the user's interest in the association rule set; thirdly, match the association rule set with the rules in the knowledge base to determine the matching result of the association rule set; then, based on the interest, matching result and association rule set, update the information tuple in the knowledge base; finally, determine the recommended course based on the information tuple in the knowledge base, and send the recommended course to the learning object. Thus, by matching the association rule set with the rules in the knowledge base and updating the information tuple in the knowledge base, dynamic and personalized course recommendations are achieved through the information tuple, which improves the reliability and accuracy of the recommended courses and enhances the learning object's experience. At the same time, the continuous updating of the knowledge base also ensures the continuous optimization and adaptability of the recommendation system.

[0011] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0013] Figure 1 is a flow chart of an embodiment of the information recommendation method according to the present disclosure;

[0014] Figure 2 It is a structural diagram of the main stages of the information recommendation method in the present disclosure;

[0015] Figure 3 It is an architecture of a client and a server disclosed in the present invention;

[0016] Figure 4 is a structural diagram of an embodiment of the information recommendation device according to the present disclosure;

[0017] Figure 5 is a structural diagram of an embodiment of the information recommendation system according to the present disclosure;

[0018] Figure 6 It is a block diagram of an electronic device used to implement the information recommendation method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0019] Unless expressly stated otherwise, throughout the specification and claims, the term "comprise" or variations such as "include" or "comprising", etc., will be understood to include the stated elements or components but not to exclude other elements or other components.

[0020] The technical solutions of the present disclosure are described below through specific examples. It should be understood that one or more steps mentioned in the present disclosure do not exclude the existence of other methods and steps before and after the combination step, or other methods and steps may be inserted between these explicitly mentioned steps. It should also be understood that these examples are only used to illustrate the present disclosure and are not used to limit the scope of the present disclosure. Unless otherwise specified, the numbering of each method step is only for the purpose of identifying each method step, and does not limit the order of arrangement of each method or limit the scope of implementation of the present disclosure. Changes or adjustments in their relative relationships can also be regarded as the scope of implementation of the present disclosure without substantial changes in the technical content.

[0021] The sources of the raw materials and instruments used in the examples are not particularly limited and can be purchased from the market or prepared according to conventional methods known to those skilled in the art.

[0022] Some current e-learning platforms have begun to utilize basic data mining and recommendation system technologies. For example, some platforms recommend new learning resources based on students' past learning records. However, these systems are generally rudimentary, limited to simple content recommendations, and lack the detailed analysis and optimization of complex data mining and collaborative filtering algorithms.

[0023] Many online education platforms collect student learning behavior data, such as course viewing time and click-through rates. However, these analyses primarily focus on descriptive statistics and lack in-depth pattern discovery. Traditional analytical methods are often static and cannot dynamically adjust teaching content.

[0024] Recommender systems have been widely used in e-commerce, primarily through collaborative filtering and association rule mining. For example, a recommendation system on a bookselling platform recommends related products based on a user's historical purchasing behavior. However, the primary goal of commercial recommendation systems is to increase sales, which differs from the goal of improving teaching effectiveness in education systems. Consequently, their data processing and recommendation rules differ.

[0025] In response to the shortcomings of traditional technologies, this paper proposes an information recommendation method, which uses an intelligent recommendation system based on association rule mining and collaborative filtering technology to provide courseware authors with specific course improvement suggestions. By continuously optimizing teaching content, it improves teaching quality and students' learning outcomes. Figure 1 A process 100 according to an embodiment of the information recommendation method of the present disclosure is shown. The information recommendation method includes the following steps:

[0026] Step 101 : determining and sending an association rule set based on a data set of a learning subject when the learning subject completes an initial course.

[0027] In this embodiment, the data set is the behavioral interaction data of the learning object, and the learning object is a student or a student with a terminal. The behavioral interaction data includes learning time, page clicks, course completion, grades, feedback, etc. The data set of the learning object completing the initial course is a multi-dimensional dynamic data set, which mainly comes from the learning object interaction behavior and learning outcome data recorded in the Learning Management System (LMS). Specifically including: the operation records of the learning object in the courseware of the initial course, such as the learning time of each unit, the page click sequence, the resource download frequency, the number of test participations, the discussion area interaction, etc.; the learning object's scores in the knowledge point tests of the initial course, the quality of homework completion, the results of periodic examinations, etc.; the learning object's ratings and comments on the initial course content, and the questions or suggestions submitted during the learning process. The data set is indirectly associated with the education object, and the education object is a teacher or a teacher with a terminal. On the one hand, the behavioral patterns in the dataset directly reflect the impact of the curriculum structure designed for the learner (such as the arrangement of knowledge points and resource distribution) on the learner's learning path. On the other hand, the teaching strategy ratings and feedback provided by the learner (such as the evaluation of the effectiveness of the teaching method for a certain knowledge point) are integrated into the dataset and used to optimize the recommendation rules. Therefore, the dataset is both an objective record of the learner's behavior and an implicit reflection of the subjective intention of the learner's instructional design.

[0028] In this embodiment, association rules are rules corresponding to interesting relationships in a data set. An association rule set includes at least one association rule. When a learning subject completes an initial course, the execution entity on which the information recommendation method runs analyzes its learning behavior and generates a set of association rule sets. These rule sets reflect the learning pattern, interests, and possible course requirements of the learning subject.

[0029] In this embodiment, the process of determining and sending a set of association rules based on the dataset of the learner upon completion of the initial course is a key step in the information recommendation method. The core of this step is to analyze the learner's behavioral data to discover potential association rules, thereby providing a basis for subsequent personalized recommendations.

[0030] Step 101 includes collecting initial data and preprocessing it to obtain a dataset. The initial data includes behavioral data on learners completing the initial course, including but not limited to course duration, course completion progress, test or assignment scores, interactive behaviors such as clicks, browsing, and saving, and feedback on the course content (e.g., ratings and comments). Data preprocessing involves cleaning, denoising, and standardizing the raw data to ensure data quality and consistency. An association rule mining algorithm is then used to mine the preprocessed initial data for frequently occurring behaviors or course combinations during the learner's learning process, generating an initial rule set. For example, if a learner completes course A, they are likely to be interested in course B as well. If a learner spends a significant amount of time on a particular knowledge point, they may need additional supplementary materials. Initial rule representation: Each association rule can be represented as ˋX→Yˋ, where ˋXˋ is the premise (e.g., having completed course A) and ˋYˋ is the conclusion (e.g., recommending course B). Based on the optimization objective, the initial rule set is filtered and optimized to obtain an association rule set. The selection criteria for selecting association rule sets from the initial rule set include support, confidence, and lift. Support is the frequency of association rules in the dataset; confidence is the credibility of the rule, that is, the probability that the conclusion is true when the premise is true; lift is the strength of the rule's relevance, measuring the independence between the premise and the conclusion. The optimization goal is to retain rules with high support, confidence, and lift, and eliminate low-quality or irrelevant rules.

[0031] In one example, the learning subject's behavioral data is as follows: they completed Course A and spent a considerable amount of time studying it. After completing Course A, they frequently browsed content related to Course B. They rated Course A highly. Through association rule mining, the following rules might be generated: Course A → Course B (support: 0.8, confidence: 0.9, lift: 1.5); Course A + high score → Course C (support: 0.7, confidence: 0.85, lift: 1.4).

[0032] In this embodiment, the initial course can be initially generated and pushed to the learner. The initial course can also be a history-recommended course. The data set also includes: feedback from the learner after completing the history-recommended course and expert verification results. The data set is mined using an association rule algorithm to obtain an association rule set. The association rule algorithm can be a Predictive Apriori algorithm.

[0033] Alternatively, association rule sets can be mined using a variety of association rule mining algorithms, including: 1) the Predictive Apriori algorithm for parameter-free association rule discovery; and 2) the IAS (Interestingness Analysis System) for subjective analysis, which classifies unexpected rules by comparing them to a previously defined knowledge database for educational domains. Educational objects with similar characteristics are discovered to be useful. Educational object characteristics are represented as a three-dimensional vector, associated with the following characteristics of their course: subject (knowledge domain, such as computer science or biology); level (course level, such as college, high school, elementary school, or special education); and difficulty (course difficulty, such as low or high). Static classification is used to compare educational objects, so similar characteristics refer to features that are completely consistent with other features. A panel of validation experts has voted on interest or effectiveness. The implemented algorithm is particularly useful in collaborative recommender systems, which can leverage the synergy provided by the network to produce increasingly useful and accurate recommendations.

[0034] Step 102: Obtain the user's interest in the association rule set.

[0035] In this embodiment, interest refers to the user's potential preference and acceptance of each association rule in the association rule set. Interest can be measured through user feedback on recommended content, click-through rate, study time, and other behavioral data. Optionally, interest can also be measured through user feedback data and interaction data. Step 102 includes collecting interactive behavior data of users interacting with the association rule set, such as: click-through rate: whether the user clicks on the course or content recommended by the association rule. Study time: the time the user spends on the recommended course corresponding to the association rule. Completion rate: whether the user completes the recommended course corresponding to the association rule. Behaviors such as collecting, sharing, and commenting. Explicit user feedback data on the data corresponding to the association rule set is used. Explicit feedback data includes ratings, likes / dislikes, and questionnaires. For example, ratings: the user's rating of the recommended content corresponding to the association rule set (e.g., 1-5 stars). Likes / dislikes: the user's expression of preference for the recommended content corresponding to the association rule set. Questionnaires: directly obtaining the user's interest in the recommended content corresponding to the association rule set through questionnaires.

[0036] Based on user behavior data, implicit interest is calculated using statistical or machine learning methods. For example, higher click-through rates indicate higher interest. Longer learning sessions indicate higher interest. Higher completion rates indicate higher interest. Based on user feedback data, interest is directly quantified. For example, higher user ratings indicate higher interest. More likes indicate higher interest. Implicit and explicit interest are combined, and weighted or other methods are used to calculate the interest level of the association rule set.

[0037] In this embodiment, as users continue to interact with recommended content, the system can update their interest in real time. For example, if a user shows a high click-through rate and study time for a newly recommended course, their interest level will be increased. Long-term updates: Regularly recalculating user interest ensures that the recommendation system can adapt to changes in user interests.

[0038] In one example, suppose the association rule set is: If the user has completed "Python Programming Basics," then "Data Science in Action" is recommended. If the user has completed "Introduction to Data Analysis," then "Machine Learning Basics" is recommended. The user clicks on the "Data Science in Action" course and studies for 30 minutes. The user rates the "Machine Learning Basics" course with 4 stars. Based on this data, the user's interest in the two recommended content is calculated: Interest in "Data Science in Action": Based on click-through rate and study time, the interest is 0.8. Interest in "Machine Learning Basics": Based on the rating, the interest is 0.9.

[0039] Step 103: Match the association rule set with the rules in the knowledge base to determine the matching result of the association rule set.

[0040] In this embodiment, the generated association rule set is matched with the existing rules in the knowledge base to determine the matching results. The matching results include: matching with the rules in the knowledge base, partially matching with the rules in the knowledge base, and not matching with any rules in the knowledge base. The matching results can help the system determine whether the current rule set is consistent with the rules in the existing knowledge base, or whether the knowledge base needs to be updated. In this embodiment, a rule matching algorithm can be used to match the association rule set with the rules in the knowledge base. Common matching processes include: exact matching, judging whether the preconditions and conclusions of the association rules are completely consistent with a rule in the knowledge base. Partial matching, judging whether the preconditions or conclusions of the association rules are partially consistent with the rules in the knowledge base (for example, the preconditions contain the same courses or knowledge points). Similarity matching, calculating the similarity between the association rules and the rules in the knowledge base (such as based on text similarity or semantic similarity), and setting a threshold to judge whether there is a match. Matching result classification: Based on the matching results, the matching results of the association rule set are divided into the following categories: complete match, the association rule is completely consistent with a rule in the knowledge base; partial match, the association rule is partially consistent with a rule in the knowledge base; new rule, the association rule does not match any rule in the knowledge base.

[0041] In this embodiment, for a complete match, if the association rule completely matches a rule in the knowledge base, the rule in the knowledge base can be used directly for recommendations without updating the knowledge base. For a partial match, if the association rule partially matches a rule in the knowledge base, the rule in the knowledge base can be updated or expanded based on user interest and the matching results. For example, the confidence level of the matching rule can be increased, or the preconditions or conclusions of the rule can be expanded. For new rules, if the association rule is completely new, it can be added to the knowledge base as a new information tuple.

[0042] Optionally, step 103 includes using a subjective analysis algorithm to match the association rule set with the rules in the knowledge base, and determining a matching result for the association rule set. A subjective analysis algorithm is a rule matching method based on domain knowledge, expert experience, or manual judgment. Unlike traditional automated matching algorithms, subjective analysis algorithms prioritize semantic similarity between rules and alignment with business logic. When rule matching requires consideration of context, semantics, or domain knowledge, subjective analysis algorithms can complement automated algorithms. The association rule set and the rules in the knowledge base are converted into a unified representation to facilitate matching. For example, the premises and conclusions of the rules are represented as structured data. The semantic information of the rules is represented as natural language text, and natural language processing (NLP) techniques are used to calculate the semantic similarity between the rules. For example, word vector models (such as Word2Vec and BERT) are used to calculate the semantic similarity between the premises and conclusions. A similarity threshold is set to determine whether the rules match. Matching is performed based on the contextual information of the rules (such as course category and learning object characteristics). For example, if the premises of two rules belong to the same course category, they are considered a partial match.

[0043] In this embodiment, in the subjective analysis stage, the execution subject on which the information recommendation method runs matches the newly discovered association rules with the existing rules in the knowledge base through the IAS algorithm (interestingness analysis system). During the matching process, the system calculates the values of four types of subjective indicators (i.e., conform_ij, unexpConseq_ij, unexpCond_ij, bsUnexp_ij) for each new rule. These values are between 0 and 1, indicating the degree of matching between the new rule and the existing rules in the knowledge base in terms of condition conformity, result unexpectedness, condition unexpectedness, or bilateral unexpectedness. For example, if the condition part of a new rule is exactly the same as that of a rule in the knowledge base, but the result part is different, then its unexpConseq_ij value may be close to 1, and it is classified as an "unexpected result rule."

[0044] Step 104: Update the information tuple in the knowledge base based on the interest level, the matching result, and the association rule set.

[0045] In this embodiment, based on the user's interest, matching results, and association rule sets, the system dynamically updates the information tuples in the knowledge base. This step ensures that the knowledge base can be continuously optimized as the user's learning behavior and interests change.

[0046] In this embodiment, an information tuple is a storage unit in the knowledge base. An information tuple includes rules, questions, recommended information, and relevance. The questions in the information tuple refer to the specific pain points in the teaching scenarios or course design corresponding to the rules discovered during the association rule mining phase. For example, through association rule analysis, it may be discovered that "after students complete Unit A, the pass rate of Unit B decreases significantly." The teaching problem reflected by this rule (such as the lack of knowledge connection between Units A and B) is abstracted as a "problem."

[0047] In this embodiment, the recommendation information in the information tuple is a specific improvement suggestion generated based on the rules and collaborative filtering results. For example, for the above-mentioned "problem", the recommendation may be "adding transitional exercises to Unit A, or adjusting the teaching order of Unit B". The generation of recommendation information can be combined with two sources: first, directly deriving possible improvement directions through IF-THEN rules. Second, combining the historical scores and feedback of education objects and experts (such as verified strategies for similar problems), recommend effective solutions recognized by peers. The lack of a "recommendation" field in the information tuple will cause the rules to remain only at the level of phenomenon description and cannot be converted into actionable course optimization measures.

[0048] In this embodiment, the correlation in the information tuple is used to quantify the degree of match between the recommendation and the question, and its determination relies on the multi-dimensional scoring in the collaborative filtering module. Specifically: the education subject votes on the practicality of the recommendation through the client to form an initial relevance score; the expert team verifies and supplements the score of the recommendation to ensure its scientificity and universality; finally, the scores of the education subject and the expert are integrated through a formula to reflect the comprehensive value of the recommended information. If the "relevance" field is missing, the system will not be able to distinguish the quality priority of the recommendation, and the education subject will need to expend extra energy to screen effective suggestions, reducing the efficiency of the system. Step 104 can dynamically optimize the knowledge base to better reflect the user's learning behavior and interest changes, thereby providing more accurate recommendations.

[0049] In this embodiment, step 104 includes: in response to the matching result being that an association rule in the association rule set completely matches a rule in the information tuple of the knowledge base, updating the recommended information and relevance in the information tuple according to the user interest and the matching result.

[0050] In this embodiment, step 104 may further include: in response to the matching result being that the association rule is completely new, generating a new information tuple according to the association rule, and adding the new information tuple to the knowledge base.

[0051] Optionally, step 104 includes: selecting an association rule with a pre-set interest level from the association rule set, performing rule matching based on the association rule using a subjective analysis algorithm, obtaining multiple types of matching results, and updating information tuples in the knowledge base by classifying the matching results. Specifically:

[0052] Conformity rule (high conform_ij value): If the new rule is highly consistent with the existing rules in the knowledge base, it will be marked as a "known valid rule" and the relevance score of the corresponding tuple in the knowledge base can be directly updated (such as increasing the voting weight) without adding a new tuple.

[0053] Unexpected result / conditional rules (high unexpConseq_ij or unexpCond_ij values): These rules reflect new causal relationships or potential problems. For example, after completing Unit A, students may not have mastered Unit B as expected, but the system finds that the completion rate of Unit C has significantly improved. In this case, the system converts such rules into new rule tuples, which contain the association rule itself, the corresponding teaching problem (such as "the association between Unit A and Unit C is not fully utilized"), and recommended information (such as "add introductory content for Unit C after Unit A"). The system also assigns an initial relevance score based on the votes of experts or education subjects.

[0054] Bilateral unexpected rules (high bsUnexp_ij values): These rules may reveal new learning patterns or design flaws and require further expert verification. The system marks them as "high-priority pending review rules" and generates temporary tuples for storage in the knowledge base. After expert verification, they will be either formalized or removed.

[0055] The specific logic behind revising the knowledge base involves determining the rule type through matching results. The system then automatically associates the new rule with the question-recommendation framework in the knowledge base, dynamically adjusts the priority of tuples (for example, giving unexpected rules a higher initial weight), and updates the relevance of tuples based on voting and prediction accuracy (using the PA algorithm). This process ensures the continuous optimization of the knowledge base and transforms subjective analysis results into actionable recommendations for course improvement.

[0056] Step 105: Determine recommended courses based on the information tuples in the knowledge base, and send the recommended courses to the learning object.

[0057] In this embodiment, the information recommendation method generates personalized course recommendations based on the information tuples in the knowledge base and sends them to the learner. The recommended courses will be more in line with the user's learning needs and interests, improving learning outcomes and user experience.

[0058] In this embodiment, the determination of recommended courses based on information tuples in the knowledge base includes extracting recommendation information of all information tuples in the knowledge base, and selecting the course with the highest frequency or number of occurrences in the recommendation information of all information tuples as the recommended course.

[0059] Optionally, after the recommended courses are sent to the learners, the learning behavior data of the learners are subsequently collected. The learning behavior data of the learners will again enter the execution entity on which the information recommendation method runs, and the knowledge base will be updated by extracting the association rule set again, forming a closed-loop iteration.

[0060] The information recommendation method provided by the embodiments of the present disclosure first determines and sends an association rule set based on the data set when the learning object completes the initial course; secondly, obtains the user's interest in the association rule set; thirdly, matches the association rule set with the rules in the knowledge base to determine the matching result of the association rule set; then, based on the interest, matching result and association rule set, updates the information tuple in the knowledge base; finally, determines the recommended course based on the information tuple in the knowledge base, and sends the recommended course to the learning object. Thus, by matching the association rule set with the rules in the knowledge base and updating the information tuple in the knowledge base, dynamic and personalized course recommendations are achieved through the information tuple, which improves the reliability and accuracy of the recommended courses and enhances the learning object's experience. At the same time, the continuous updating of the knowledge base also ensures the continuous optimization and adaptability of the recommendation system.

[0061] In a specific example, Figure 2 As shown, the main stages of the information recommendation method disclosed in the present invention include: the first stage to the fourth stage. In the first stage to the fourth stage, each stage plays its own role and finally obtains the recommended courses.

[0062] The first stage is association rule set mining: The goal of this stage is to find association rule sets on the dataset generated when the learning subject completes the course. Once the data is preprocessed, the output of this stage (i.e., the determined association rule set) then enters the second stage for subjective analysis. This stage uses subjective rule evaluation metrics to determine the interestingness of the rules found through association rule set mining. It also applies the IAS algorithm to compare these rules with the rules in the knowledge base to classify them as expected or unexpected. Among them, subjective rule evaluation metrics refer to evaluation methods that use the professional knowledge or experience of users (such as education subjects, education experts) to make value judgments on the rules generated by association rule mining. Its core lies in combining the user's subjective cognition rather than relying solely on objective statistical indicators (such as support and confidence). Specifically, the interestingness of a rule represents the degree of attention and practicality evaluation of the rule by the courseware author or education expert. For example, a rule may show that "learning subjects who complete unit A have an 80% probability of mastering unit B." If the rule is consistent with the teaching experience of the education subject, it is classified as "rule-compliant" and has a high level of interest. If the rule reveals an association that the education subject did not expect (such as "learning subjects who frequently use a certain interactive tool have significantly improved their grades"), it may be marked as an "unexpected rule" and has a high level of interest because it provides a new optimization direction.

[0063] "Expected" rules refer to new rules discovered through association rule mining that are highly consistent with existing rules in the knowledge base or prior knowledge in the education field. This means that the content, logical relationships, or application scenarios of these rules align with the existing experience of the learners and education experts, or with proven teaching strategies. For example, if the knowledge base already contains the rule "After the learner completes Unit A, the pass rate of Unit B significantly increases," and the newly discovered rule "After the learner completes Unit A, the completion time of Unit B is shortened" has a strong correlation between the condition and the result, it may be classified as "expected."

[0064] "Unexpected" rules refer to the creation of knowledge within the knowledge base, which is a "one-time initialization, continuous optimization" process. Each time the system runs, it incrementally updates the existing knowledge base, incorporating new data and feedback, rather than rebuilding from scratch. This design ensures the stability of the knowledge base and the reuse of historical experience, while also enabling dynamic evolution through collaborative mechanisms, ultimately forming an intelligent rule base with self-improvement capabilities.

[0065] The third phase is the knowledge base update phase: This phase combines collaborative filtering and knowledge-based techniques to update the rule base. The knowledge base is stored in the form of information tuples (rule-question-recommendation-relevance) and categorized according to specific course profiles. To avoid the cold start problem of collaborative filtering systems, experts propose the first information tuple in the repository and vote on tuples proposed by other experts. On the other hand, education subjects can discover new information tuples, which are verified by expert subjects before being inserted into the repository and vote on other tuples.

[0066] The fourth phase is the recommendation phase. The expected rules found in the second phase are combined with the more intuitive information tuples from the third phase. These are then used in the fourth phase to present possible solutions to problems detected in the course to the learners, who are generally not skilled in data mining. The learners analyze the recommendations and determine their relevance, ultimately deciding on the recommended course.

[0067] Among them, the recommendation information for the fourth stage mainly includes: for example, "The interaction rate of Unit C is lower than the average level, which may affect the mastery of knowledge points." It is generated based on association rules and recommendation fields in tuples, such as "It is recommended to add a simulation experiment module to Unit C (from rule R001)" and "Refer to the interactive question-and-answer strategies adopted by other educational objects in similar courses (collaborative filtering recommendation)." The support, confidence, and voting ratio of educational objects / experts of the rules are displayed in a visual form, such as "This suggestion has been adopted by 85% of educational objects in similar courses" and "Expert rating 4.8 / 5.0." It provides modification options that can be directly embedded in courseware (such as inserting specific resource links and adjusting the order of chapters), and supports one-click application or further customization.

[0068] In some embodiments of the present disclosure, the above-mentioned information recommendation method also includes: obtaining a data set of recommended courses completed by the learning object; determining and sending a new association rule set based on the data set of recommended courses completed by the learning object; generating new recommended courses based on the new association rule set, the user's interest in the new association rule set, and the matching results of the new association rule set in the knowledge base, and sending the new recommended courses to the learning object.

[0069] In this embodiment, the dataset of the learning subject's completion of the recommended course includes: the learning subject's operation records in the courseware of the recommended course, the learning subject's scores in the quizzes on various knowledge points of the recommended course, the quality of homework completion, the results of periodic examinations, etc., the learning subject's ratings and comments on the recommended course content, and questions or suggestions submitted during the learning process.

[0070] In this embodiment, the above-mentioned determination and sending of a new association rule set based on the data set of recommended courses completed by the learning object includes: selecting the evaluation results of the learning object, such as scores, comments, questions or suggestions, from the data set of recommended courses completed by the learning object, and updating the association rule set based on the evaluation results to obtain a new association rule set.

[0071] In this embodiment, the method for obtaining the user's interest in the new association rule set can refer to the above-mentioned process of obtaining the user's interest in the association rule set, and the process of obtaining the matching result of the new association rule set in the knowledge base can refer to the process of obtaining the matching result of the association rule set in the knowledge base.

[0072] The information recommendation method provided in this embodiment obtains a data set of recommended courses completed by a learning object; determines and sends a new association rule set based on the data set of recommended courses completed by the learning object; generates new recommended courses based on the new association rule set, the user's interest in the new association rule set, and the matching results of the new association rule set in the knowledge base, and sends the new recommended courses to the learning object, thereby ensuring iterative optimization of information in the knowledge base and recommended information, and improving the accuracy and reliability of the new recommended courses.

[0073] In some embodiments of the present disclosure, the above-mentioned determination and sending of the association rule set based on the data set when the learning object completes the initial course includes: preprocessing the data set when the learning object completes the initial course to obtain preprocessed data; using an association analysis algorithm to perform rule analysis on the preprocessed data and the search area selected by the education object to obtain an association rule set; and sending the association rule set to the education object.

[0074] In this embodiment, the preprocessing mentioned above involves data cleaning and normalization of the dataset. This preprocessing allows the preprocessed data to be suitable for association rule analysis. The search region is the data range within the preprocessed data determined by the educational target, such as data specificity and data attributes. Selecting the search region by the educational target can help streamline the targeting area.

[0075] In this embodiment, an association rule set is determined and sent based on the dataset generated when the learner completes the initial course. The specific implementation steps and methods are as follows: Step 1: Data preprocessing: The dataset generated when the learner completes the initial course is cleaned and organized to make it suitable for subsequent association rule analysis. Step 2: Association rule analysis: An association analysis algorithm is used to analyze the data and attributes in the search area selected by the learner in the preprocessed data to identify association rules between courses. The generated association rule set is sent to the learner (e.g., the learner, administrator, or recommendation system).

[0076] The method for determining and sending an association rule set provided by the present disclosure preprocesses the dataset of a learning subject when they complete an initial course to obtain preprocessed data; uses an association analysis algorithm to perform rule analysis on the preprocessed data and a search area selected by the learning subject to obtain an association rule set; and then sends the association rule set to the learning subject. This improves the accuracy and reliability of the association rule set.

[0077] Optionally, the above-mentioned determination and sending of the association rule set based on the data set when the learning object completes the initial course includes: preprocessing the data set when the learning object completes the initial course to obtain preprocessed data; using an association analysis algorithm to perform rule analysis on the preprocessed data to obtain an association rule set; and sending the association rule set to the education object.

[0078] In some optional implementations of the present disclosure, the above-mentioned obtaining of the user's interest in the association rule set includes: obtaining user feedback data on each association rule in the association rule set; based on the feedback data, determining the first voting ratio of the education subjects for each association rule and the second voting ratio of the expert subjects for each association rule; based on the first voting ratio and the second voting ratio, determining the interest level of each association rule in the association rule set.

[0079] In this optional implementation, users include educational objects and expert objects. The educational object is an object that outputs recommended courses to the learning object. For example, the educational object is an educational object or an educational object with a smart terminal. After obtaining the association rule set, the educational object evaluates the association rules in the association rule set and gives feedback. The expert object is an object that comprehensively evaluates the educational object and the learning object. For example, the expert object is an expert or an expert with a smart terminal. After obtaining the association rule set, the expert object evaluates the association rules in the association rule set and gives feedback. The feedback from the educational object and the feedback from the expert object are used as the user's feedback data.

[0080] In this optional implementation, user feedback data on each association rule in the association rule set includes providing a feedback interface or tool for users to evaluate the association rule, such as providing options of "agree," "disagree," or "neutral," and providing a rating function (e.g., 15 stars).

[0081] In this optional implementation, the first votes of education subjects and the second votes of expert subjects in the feedback data are distinguished. The number of first votes and the number of second votes are taken as the total number of votes. The first vote percentage is equal to the number of first votes for each association rule divided by the total number of votes, and the second vote percentage is equal to the number of second votes for each association rule divided by the total number of votes. The first vote percentage and the second vote percentage of each association rule are weighted and summed to obtain the interest level of each association rule. The weights of the first vote percentage and the second vote percentage can be set based on needs, such as the weight of the second vote percentage being greater than the weight of the first vote percentage.

[0082] Specifically, the weighted sum of the first and second voting percentages for each association rule to obtain the interest level for each association rule includes the following: The weighted sum of the first and second voting percentages is used to obtain the interest level. For example, interest level = w1 × first voting percentage + w2 × second voting percentage; where w1 and w2 are weight parameters, such that w1 + w2 = 1. Weights can be adjusted based on business needs. If expert opinions are given greater weight, w2 can be set to 0.7 and w1 to 0.3.

[0083] This optional implementation provides a method for determining the interest level of each association rule in an association rule set, obtaining user feedback data on each association rule in the association rule set; based on the feedback data, determining the first voting percentage of the education subjects for each association rule and the second voting percentage of the expert subjects for each association rule; based on the first voting percentage and the second voting percentage, determining the interest level of each association rule in the association rule set, thereby improving the reliability and accuracy of obtaining the interest level.

[0084] Optionally, the above-mentioned obtaining of the user's interest in the association rule set includes: obtaining user feedback data on each association rule in the association rule set; determining the association rule set recognized by the user based on semantic information of the feedback data, and determining the recognition degree value of the association rule based on the semantic information, and using the recognition degree value of each association rule as the interest degree of each association rule in the association rule set.

[0085] In some optional implementations of the present disclosure, the above-mentioned determination of the interest level of each association rule in the association rule set based on the first voting ratio and the second voting ratio includes: determining the first weighting coefficient of the education object and the second weighting coefficient of the expert object; obtaining a weighted measurement value based on the first voting ratio, the second voting ratio, the first weighting coefficient, and the second weighting coefficient; predicting the accuracy of the voting results of the education object on each association rule to obtain the accuracy results of each association rule; and obtaining the interest level of each association rule based on the weighted measurement value and the accuracy result.

[0086] In this optional implementation, let U1, U2, ..., Um For m different educational objects with different data sources, S i For U i (i=1,2,...,m) the association rule set found, S={S1,S2,...,S m}; Let E1, E2, ..., E k are k different experts. According to the weight definition, the votes on the rule R in S can be used to assign weight W to R. R .

[0087] In fact, the education subjects tend to apply those association rules that have more support or more votes, and the first vote ratio is shown in formula (1).

[0088]

[0089] In formula (1), i = 1, 2, ..., n, NumVotesTeachers (R i ) is the rule R in S i The number of education subjects who voted.

[0090] By applying the same reasoning to expert object voting, the second voting proportion is shown in formula (2):

[0091]

[0092] In formula (2), i=1, 2, ... n, NumVotesExperts(R i ) is the rule R in R i The number of experts who voted.

[0093] Therefore, Rule R i The weight of can be expressed as a weighted measure of the votes of education objects and expert objects, that is:

[0094]

[0095] In formula (3), C u and C e are the first weighting coefficient and the second weighting coefficient representing the education object and the expert object respectively.

[0096] Once the weight of each rule is calculated, an interest metric can be designed, also known as weighted accuracy. According to the PA algorithm, the prediction accuracy of the rule can be calculated as i Defining Interest As shown in formula (4):

[0097]

[0098] in is the weighted metric calculated according to formula (3), accR ij The PA algorithm is used to vote for rule R i The prediction accuracy result returned by the education object.

[0099] In this embodiment, the PA algorithm (Predictive Apriori Algorithm) provides a prediction accuracy result for each rule. This accuracy result, calculated by the PA algorithm using Bayesian methods, reflects the rule's ability to predict future data and is used to comprehensively measure the rule's effectiveness and practicality. By using prediction accuracy as an important basis for ranking and recommending rules, it ultimately influences the priority and credibility of the system's recommendations to courseware authors.

[0100] In some optional implementations of the present disclosure, the above-mentioned updating of information tuples in the knowledge base based on interest, matching results and association rule sets includes: based on interest, filtering the rule set to be converted from the association rule set; filtering the matching results related to the rule set to be converted from the matching results; based on the matching results of the rule set to be converted, converting the rule set to be converted into an information tuple; and using the converted information tuple to update the information tuple in the knowledge base.

[0101] In this optional implementation, the association rule with the highest interest is determined, the selected association rule is used as the rule set to be converted, and the rule set to be converted is screened from the association rule set. The above-mentioned screening of matching results related to the rule set to be converted from the matching results includes: based on the association rules in the rule set to be converted, using the matching results related to the association rules as the matching results related to the rule set to be converted.

[0102] In this optional implementation, relevant matching results are filtered out from the matching results based on the rule set to be converted. For example: if the rule set to be converted contains rule A, the matching results of rule A are filtered out. Information tuples are usually represented in the form of structured data, such as: rules, questions, recommended information, and relevance. Conformity: If the matching result is "conformity", the information tuple in the knowledge base is used directly without conversion. Unexpected result: If the matching result is "unexpected result", a new information tuple is generated, the premise is retained, and the conclusion is updated. Unexpected condition: If the matching result is "unexpected condition", a new information tuple is generated, the conclusion is retained, and the premise is updated. Bilateral surprise: If the matching result is "bilateral surprise", a brand new information tuple is generated.

[0103] In this optional implementation, converting the rule set to be converted into information tuples based on the matching results of the rule set to be converted includes: generating information related to the rules in the rule set to be converted, such as teaching questions, recommendations, and relevance scores. For example, if the rule is "Unit A → Unit B," the corresponding teaching question is "Low completion rate for Unit B"; the recommendation is "Additional exercises after Unit A is recommended"; and a relevance score is generated based on voting and prediction accuracy.

[0104] In this optional implementation, the above-mentioned use of the converted information tuples to update the information tuples in the knowledge base includes: classifying the converted information tuples according to course characteristics (topic, difficulty, level) and storing them in the knowledge base so that the information tuples in the knowledge base support collaborative filtering matching.

[0105] Optionally, using the converted information tuple to update the information tuple in the knowledge base further includes: adding the converted information tuple to the knowledge base and removing duplicate information tuples. Inserting the converted information tuple into the knowledge base. Checking whether there is a rule in the knowledge base that is identical to the newly added information tuple (e.g., the premise and conclusion are completely consistent), and if so, removing the duplicate information tuple.

[0106] The method for updating information tuples in the knowledge base provided by the present disclosure can dynamically update the information tuples in the knowledge base based on interest, matching results and association rule sets, thereby providing more accurate support for subsequent personalized recommendations.

[0107] In some optional implementations of the present disclosure, the above-mentioned use of converted information tuples to update the information tuples in the knowledge base includes: adding the converted information tuples to the knowledge base and removing duplicate information tuples in the knowledge base; collecting the prediction accuracy of each information tuple in the knowledge base; determining the relevance of each information tuple; determining the priority of each information tuple based on the prediction accuracy and relevance; and sorting the information tuples in the knowledge base based on the priority of each information tuple.

[0108] In this optional implementation, the prediction accuracy refers to the recommendation accuracy of the rule in actual application, which is usually calculated by the following formula: PA = number of correct recommendations / total number of recommendations, where the number of correct recommendations and the total number of recommendations for each rule are counted through historical recommendation data or experimental data.

[0109] Optionally, the predicted accuracy of each information tuple in the collected knowledge base includes predicting the predicted accuracy of each information tuple using the PA (Apriori) algorithm, where the PA algorithm provides an accurate predicted accuracy for each rule. This predicted accuracy, calculated by the PA algorithm using a Bayesian approach, reflects the rule's ability to predict future data. The predicted accuracy generated by the PA algorithm is incorporated into the comprehensive evaluation of the rule, ultimately influencing the priority and credibility of the system's recommendations to courseware authors.

[0110] Optionally, the prediction accuracy can also be combined with the voting proportions of education objects and expert objects to form a weighted accuracy rate (WAcc), which is used to comprehensively measure the effectiveness and practicality of the rules.

[0111] In this optional implementation, the relevance of each information tuple in the knowledge base is determined to assess the importance of the rule. Relevance refers to the degree to which the rule fits the current learning object or business requirement. The relevance score for each rule can be calculated based on the frequency of use of the rule (e.g., the number of recommendations), the interest level of the rule (e.g., user feedback data), and the semantic relevance of the rule (e.g., the degree of match with the current learning object).

[0112] In this optional implementation, determining the priority of each information tuple based on prediction accuracy and relevance includes: performing a weighted summation of the prediction accuracy and relevance to obtain a priority score. Normalization: Normalizing the priority score to between 0 and 1 to facilitate subsequent sorting. Calculating the priority of each information tuple and sorting the information tuples in the knowledge base. Sort the information tuples from high to low according to the priority score. Updating the sorted information tuples to the knowledge base to ensure that high-priority rules are used first.

[0113] The method for updating information tuples in the knowledge base provided by this optional implementation can use the converted information tuples to update the information tuples in the knowledge base, and sort the rules based on prediction accuracy and relevance, thereby providing more accurate support for subsequent personalized recommendations.

[0114] In some optional implementations of the present disclosure, the information tuples in the above-mentioned knowledge base are constructed through the following steps: obtaining the cold start rules of the expert object; determining the expert tuple based on the cold start rules; obtaining the education tuple sent by the education object; sending the education tuple to the expert object; in response to receiving the verification result of the education tuple by the expert object, using both the expert tuple and the education tuple as information tuples in the knowledge base.

[0115] In this optional implementation, the cold start rule is a rule pre-set by the expert object for the knowledge base during cold start. The above-mentioned determination of the expert tuple based on the cold start rule includes: determining the problems in the teaching scenario or course design based on the cold start rule; determining the recommended information based on the problem, and sending the problem and the recommended information; obtaining the correlation between the problem and the recommended information; and using the cold start rule, problem, recommended information and correlation as the expert tuple.

[0116] In this optional implementation, the education tuple is a tuple whose education object is a cold start setting of the knowledge base. The structure of the education tuple can be: rules, questions, recommendation information and relevance.

[0117] In this alternative implementation, to avoid the cold start problem of collaborative filtering systems, experts propose the first tuple in the repository and vote on tuples proposed by other experts. Teachers, on the other hand, can discover new tuples, which must be verified by experts before being inserted into the repository, and vote on other tuples, improving the reliability of cold starts in the knowledge base.

[0118] In some optional implementations of the present disclosure, the above-mentioned information tuple is constructed through the following steps: determining association rules; based on the association rules, determining problems in the teaching scenario or course design; based on the problems, determining recommended information, and sending the problems and recommended information; obtaining the correlation between the problems and the recommended information; and using the association rules, problems, recommended information and correlation as information tuples.

[0119] In this optional implementation, association rules can be extracted from the behavioral data of learning subjects using association rule mining algorithms (such as Apriori and FPGrowth). The specific steps for extracting association rules include: collecting learning subject behavioral data (such as course completion records, test scores, and study time). Using the association rule mining algorithm, a set of association rules is generated. Each rule is typically expressed in the form of "IFTHEN." For example, if a user has learned "Python Programming Basics," then "Data Science Practice" is recommended.

[0120] In this optional implementation, based on association rules, problems in teaching scenarios or course design are identified, including:

[0121] Identify potential teaching problems through the premises and conclusions of association rules. For example, if a user does not complete "Data Science Practice" after learning "Python Programming Basics", there may be a course connection problem. Classify the identified problems (such as course design problems, learning path problems, knowledge point coverage problems, etc.).

[0122] In this optional implementation, determining recommendation information based on the question includes generating recommendation information based on question type and association rules. For example, if there is a course connection issue, "Data Science in Action" may be recommended as a follow-up course; if there is a knowledge point coverage issue, supplementary learning materials may be recommended. The recommendation information is converted into structured data (e.g., JSON or XML).

[0123] In this optional implementation, relevance refers to the degree of fit between the recommended information and the question. The relevance score between the recommended information and the question can be calculated based on user feedback data (such as ratings, click-through rates) and the actual effects of the recommended information (such as the completion rate of the learning object, test scores).

[0124] The method for constructing information tuples provided in the present disclosure can construct information tuples and use them for updating the knowledge base and personalized recommendations, thereby improving teaching effects and learning experience.

[0125] In some optional implementations of the present disclosure, the above-mentioned determining recommended courses based on information tuples in the knowledge base and sending recommended courses to the learning object includes: determining the ranking of each information tuple based on the information tuples in the knowledge base; determining the information tuple to be recommended based on the ranking; extracting the recommendation information in the information tuple to be recommended, and sending the recommendation information to the education object; receiving the recommended courses sent by the education object, and sending the recommended courses to the learning object.

[0126] In this optional implementation, determining recommended courses based on information tuples in the knowledge base includes sorting the information tuples in the knowledge base according to ranking metrics such as priority, interest, and confidence. Priority is calculated based on prediction accuracy and relevance, interest is calculated based on user feedback data, and confidence is calculated based on the confidence level of the rule. When sorting the information tuples, the information tuples can be sorted from high to low according to the ranking metrics. For example, information tuples with higher priorities are ranked first.

[0127] In this optional implementation, the above-mentioned ranking-based determination of information tuples to be recommended includes: setting screening conditions according to business requirements, wherein the screening conditions include: selecting the top N information tuples with the highest priority, and selecting information tuples with interest greater than a certain threshold.

[0128] In this optional implementation, the information tuple to be recommended includes multiple recommendation information. Sending the recommendation information to the education object can enable the education object to recommend the recommended course that is most relevant to the multiple recommendation information, thereby facilitating sending the recommended course to the learning object.

[0129] The method for determining recommended courses provided by this optional implementation can determine recommended courses based on information tuples in the knowledge base and send them to the learning object, thereby improving the learning experience and effect. The information recommendation method provided by this disclosure is based on a client-server architecture. Figure 3 As shown, there are multiple clients that apply association rule mining algorithms locally on the data of online courses used by learning objects. The server application consists of two modules. The first is a web application server (for installing Figure 3 ) so that expert Z can manage the knowledge base (KB), add, delete or edit tuples, and vote on the contributions made by other experts in the team. The second module is a web service that allows the server to share updated KBs with clients in PMML format. PMML (Predictive Model Markup Language) is an XML-based language that makes it possible to define and share predictive models between different applications, establishing a vendor-independent way of defining these models, thus avoiding proprietary applications and compatibility issues. Therefore, once an updated version of the KB has been downloaded from the server, the client can apply the mining algorithm offline. The client application is part of the iterative methodology used by educational subjects to develop courses. It is able to detect possible problems in the design and content of e-learning courses by adding feedback or maintenance phases to the courses.

[0130] like Figure 3 As shown, there are several stages in this methodology: 1) Initial construction of the course, e.g. Figure 3 1) The educational object J creates or modifies the course; 2) The learning object X completes the course, i.e. Figure 3 During the course, usage information is transparently compiled and stored in the knowledge base ( Figure 3 The data used in the process is the data in the knowledge base); 3) the continuous improvement phase, which is carried out in parallel with the client application. The last phase contains the core of the rule mining algorithm used. The algorithm, together with the knowledge base, classifies the found rules as expected (if consistent with the knowledge base) or unexpected (if inconsistent), corresponding to Figure 3 "Recommendations" and "surprises" in [1]. If educator J applies a recommendation to a course, they implicitly vote on its usefulness in the server knowledge base. Unexpected tuples are ranked according to the IAS algorithm, and educators can mark any interesting tuples. Expert Z then analyzes these unexpected "interesting" or "valid" information tuples and may choose to include them in the knowledge base.

[0131] Further references Figure 4 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of an information recommendation device. Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0132] like Figure 4 As shown, the information recommendation device 400 provided in this embodiment includes: a determination unit 401, an acquisition unit 402, a matching unit 403, an update unit 404, and a recommendation unit 405. The determination unit 401 can be configured to determine and send an association rule set based on the dataset of the learning subject when the learning subject completes the initial course. The acquisition unit 402 can be configured to obtain the user's interest in the association rule set. The matching unit 403 can be configured to match the association rule set with the rules in the knowledge base and determine the matching result of the association rule set. The updating unit 404 can be configured to update the information tuple in the knowledge base based on the interest, the matching result, and the association rule set. The recommendation unit 405 can be configured to determine recommended courses based on the information tuple in the knowledge base and send the recommended courses to the learning subject.

[0133] In this embodiment, the specific processing of the determination unit 401, the acquisition unit 402, the matching unit 403, the updating unit 404, and the recommendation unit 405 and the technical effects thereof can be referred to in detail. Figure 1 The relevant descriptions of step 101, step 102, step 103, step 104, and step 105 in the corresponding embodiment are not repeated here.

[0134] In some embodiments of the present disclosure, the above-mentioned determination unit 401 is further configured to: preprocess the data set when the learning object completes the initial course to obtain preprocessed data; use an association analysis algorithm to perform rule analysis on the preprocessed data and the search area selected by the education object to obtain an association rule set; and send the association rule set to the education object.

[0135] In some embodiments of the present disclosure, the acquisition unit 402 is further configured to: acquire user feedback data on each association rule in the association rule set; determine the first voting ratio of the education subjects for each association rule and the second voting ratio of the expert subjects for each association rule based on the feedback data; and determine the interest level of each association rule in the association rule set based on the first voting ratio and the second voting ratio.

[0136] In some embodiments of the present disclosure, the above-mentioned update unit 404 is configured to: filter the rule set to be converted from the association rule set based on the interest level; filter the matching results related to the rule set to be converted from the matching results; convert the rule set to be converted into an information tuple based on the matching results of the rule set to be converted; and use the converted information tuple to update the information tuple in the knowledge base.

[0137] In some embodiments of the present disclosure, the above-mentioned update unit 404 is further configured to: add the converted information tuples to the knowledge base and remove duplicate information tuples in the knowledge base; collect the prediction accuracy of each information tuple in the knowledge base; determine the relevance of each information tuple; determine the priority of each information tuple based on the prediction accuracy and relevance; and sort the information tuples in the knowledge base based on the priority of each information tuple.

[0138] In some embodiments of the present disclosure, the above-mentioned device also includes: a construction unit (not shown in the figure) for constructing an information tuple, and the above-mentioned construction unit is configured to: determine association rules; determine problems in the teaching scenario or course design based on the association rules; determine recommended information based on the problems, and send the problems and recommended information; obtain the correlation between the problems and the recommended information; and use the association rules, problems, recommended information and correlation as information tuples.

[0139] In some embodiments of the present disclosure, the above-mentioned recommendation unit 405 is further configured to: determine the ranking of each information tuple based on the information tuples in the knowledge base; determine the information tuple to be recommended based on the ranking; extract the recommendation information in the information tuple to be recommended, and send the recommendation information to the education object; receive the recommended courses sent by the education object, and send the recommended courses to the learning object.

[0140] The information recommendation device provided by the embodiments of the present disclosure includes: first, a determination unit 401 determines and sends an association rule set based on the dataset of the learner when completing the initial course; second, an acquisition unit 402 obtains the user's interest in the association rule set; third, a matching unit 403 matches the association rule set with the rules in the knowledge base and determines the matching result of the association rule set; then, an updating unit 404 updates the information tuple in the knowledge base based on the interest, the matching result, and the association rule set; finally, a recommendation unit 405 determines a recommended course based on the information tuple in the knowledge base and sends the recommended course to the learner. Thus, by matching the association rule set with the rules in the knowledge base and updating the information tuple in the knowledge base, dynamic and personalized course recommendations are achieved through the information tuple, improving the reliability and accuracy of the recommended courses and enhancing the learner's experience. At the same time, the continuous updating of the knowledge base also ensures the continuous optimization and adaptability of the recommendation system.

[0141] Further references Figure 5 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of an information recommendation system. Figure 1 The method embodiment shown corresponds to the embodiment shown.

[0142] like Figure 5 As shown, the information recommendation system 500 provided in this embodiment includes: a client 501 and a server 502.

[0143] The client 501 is used to determine and send an association rule set based on the dataset of the learning subject when completing the initial course; obtain the user's interest in the association rule set; match the association rule set with the rules in the knowledge base 5011 to determine the matching result of the association rule set; update the information tuples in the knowledge base 5011 based on the interest, the matching result, and the association rule set; and determine recommended courses based on the information tuples in the knowledge base 5011 and send the recommended courses to the learning subject. The client 501 can be a terminal used by an educational subject (such as a teacher, training subject, etc.). The client locally applies an association rule mining algorithm as part of the iterative course development process and can be used to identify and provide feedback. The problem and feedback can be directed to the subject being educated by the educational subject, such as the learning subject. The problem can be a recommended course for the learning subject. The client can download the knowledge base, receive updates to the knowledge base, and apply the mining algorithm to the knowledge base offline to update the knowledge base.

[0144] The server 502 includes a network application module 5021 and a network service module 5022. The network application module 5021 provides an interface for managing the knowledge base 5011, allowing the knowledge base 5011 to be updated via the interface. The network service module 5022 provides the updated knowledge base to the client 501. The interface provided by the network application module 5021 allows experts to manage the knowledge base through methods such as addition, deletion, modification, query, and voting. Experts use the interface to operate the knowledge base to achieve the purpose of linking knowledge bases. After receiving the updated knowledge base, the client 501 performs the same management on the local knowledge base 5011 as it did on the received updated knowledge base.

[0145] In this embodiment, the network service module 5022 uses PMML (Predictive Model Markup Language) to share the knowledge base updated by the network application module 5021 with the client 501. The use of PMML in the network service module 5022 ensures the standardization and interoperability of model sharing and avoids compatibility issues.

[0146] In this embodiment, in the information recommendation system 500, the specific processing of the client 501 and the technical effects thereof can be referred to in Figure 1 The relevant descriptions of step 101, step 102, step 103, step 104, and step 105 in the corresponding embodiment are not repeated here.

[0147] The information recommendation system provided in this embodiment discovers patterns through client-side data mining, evaluates and refines the knowledge base based on the knowledge of experts and teachers, and finally feeds back recommended courses to learners in the form of recommendations based on the knowledge of the updated knowledge base, forming a closed loop and achieving continuous improvement of e-learning courses.

[0148] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0149] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their modes are provided for example only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0150] like Figure 6 As shown, the device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. Various programs and data required for the operation of the device 600 can also be stored in the RAM 603. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0151] Various components in device 600 are connected to I / O interface 605, including an input unit 606, such as a keyboard, mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a magnetic disk, optical disk, etc.; and a communication unit 609, such as a network card, modem, wireless communication transceiver, etc. The communication unit 609 allows device 600 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0152] The computing unit 601 can be a variety of general-purpose and / or specialized processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the information recommendation method. For example, in some embodiments, the information recommendation method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the information recommendation method described above can be performed. Alternatively, in other embodiments, the computing unit 601 can be configured to perform the information recommendation method by any other appropriate means (e.g., by means of firmware).

[0153] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0154] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable information recommendation device, so that when the program code is executed by the processor or controller, the modes / operations specified in the flowcharts and / or block diagrams are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0155] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0156] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0157] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0158] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0159] The foregoing descriptions of specific exemplary embodiments of the present disclosure are for purposes of illustration and description. These descriptions are not intended to limit the present disclosure to the precise forms disclosed, and it is apparent that many variations and modifications are possible in light of the foregoing teachings. The exemplary embodiments have been selected and described for the purpose of explaining the specific principles of the present disclosure and their practical application, thereby enabling those skilled in the art to realize and utilize a variety of exemplary embodiments of the present disclosure and various options and modifications. The scope of the present disclosure is intended to be defined by the claims and their equivalents.

Claims

1. An information recommendation method, comprising: Determine and send a set of association rules based on the data set when the learning object completes the initial course; Obtaining the user's interest in the association rule set; Matching the association rule set with the rules in the knowledge base to determine a matching result of the association rule set; Based on the interest level, the matching result and the association rule set, updating the information tuple in the knowledge base; Based on the information tuples in the knowledge base, recommended courses are determined, and the recommended courses are sent to the learning object.

2. The method according to claim 1, wherein The method further comprises: Obtaining a data set of the learning subject completing the recommended course; Determining and sending a new set of association rules based on a data set of the learning object completing the recommended course; Based on the new association rule set, the user's interest in the new association rule set, and the matching result of the new association rule set in the knowledge base, a new recommended course is generated, and the new recommended course is sent to the learning object.

3. The method according to claim 1, wherein The determining and sending of the association rule set based on the data set when the learning object completes the initial course includes: Preprocessing the data set when the learning object completes the initial course to obtain preprocessed data; Using an association analysis algorithm to perform rule analysis on the pre-processed data and the search area selected by the education object to obtain an association rule set; The association rule set is sent to the education object.

4. The method according to claim 1, wherein The obtaining of the user's interest in the association rule set includes: Obtaining user feedback data on each association rule in the association rule set; Determining, based on the feedback data, a first voting percentage of the education subjects for each association rule and a second voting percentage of the expert subjects for each association rule; The interest level of each association rule in the association rule set is determined based on the first voting percentage and the second voting percentage.

5. The method according to claim 4, wherein The determining, based on the first voting proportion and the second voting proportion, the interest level of each association rule in the association rule set includes: determining a first weighting coefficient of the education object and a second weighting coefficient of the expert object; Obtaining a weighted metric value based on the first voting proportion, the second voting proportion, the first weighting coefficient, and the second weighting coefficient; Predicting the accuracy of the voting results of the education subjects on each association rule to obtain the accuracy results of each association rule; Based on the weighted metric value and the accuracy result, the interest level of each association rule is obtained.

6. The method according to claim 1, wherein The updating of the information tuple in the knowledge base based on the interest level, the matching result and the association rule set includes: Based on the interest level, selecting a rule set to be converted from the association rule set; Filtering matching results related to the rule set to be converted from the matching results; Based on the matching result of the rule set to be converted, converting the rule set to be converted into an information tuple; The converted information tuple is used to update the information tuple in the knowledge base.

7. The method according to claim 6, wherein: The updating of the information tuple in the knowledge base using the converted information tuple comprises: Adding the converted information tuples to the knowledge base and removing duplicate information tuples in the knowledge base; Collecting the prediction accuracy of each information tuple in the knowledge base; Determine the relevance of each information tuple; Determining the priority of each information tuple based on the prediction accuracy and the correlation; The information tuples in the knowledge base are sorted based on the priority of each information tuple.

8. The method according to any one of claims 1 to 5, wherein: The information tuples in the knowledge base are constructed by the following steps: Get the cold start rules of the expert object; Determining an expert tuple based on the cold start rule; Get the education tuple sent by the education object; sending the education tuple to the expert object; In response to receiving a verification result of the education tuple by the expert object, both the expert tuple and the education tuple are used as information tuples in the knowledge base.

9. The method according to any one of claims 1 to 5, wherein: The information tuples in the knowledge base can also be constructed by the following steps: Determine association rules; Based on the association rules, identify problems in the teaching scenario or course design; Determining recommended information based on the question, and sending the question and the recommended information; Obtaining the relevance between the question and the recommended information; The association rules, the questions, the recommendation information, and the relevance are taken as information tuples.

10. The method according to claim 1, wherein The determining of recommended courses based on the information tuples in the knowledge base and sending the recommended courses to the learning object includes: Determining the order of each information tuple based on the information tuples in the knowledge base; Based on the ranking, determining the information tuple to be recommended; Extracting the recommended information from the information tuple to be recommended, and sending the recommended information to the education object; Receive the recommended courses sent by the education object, and send the recommended courses to the learning object.

11. An information recommendation device, comprising: a determining unit configured to determine and send an association rule set based on a data set when the learning object completes an initial course; an acquisition unit configured to acquire a user's interest in the association rule set; a matching unit configured to match the association rule set with rules in a knowledge base and determine a matching result of the association rule set; an updating unit, configured to update the information tuple in the knowledge base based on the interest level, the matching result and the association rule set; The recommendation unit is configured to determine recommended courses based on the information tuples in the knowledge base and send the recommended courses to the learning object.

12. An information recommendation system, comprising: Client and server; The client is used to determine and send an association rule set based on the data set when the learning object completes the initial course; Obtaining the user's interest in the association rule set; Matching the association rule set with the rules in the knowledge base to determine a matching result of the association rule set; Based on the interest level, the matching result and the association rule set, updating the information tuple in the knowledge base; Determining recommended courses based on the information tuples in the knowledge base, and sending the recommended courses to the learning object; The server includes: a network application module and a network service module, wherein the network application module is used to provide an interface for managing the knowledge base, so as to update the knowledge base through the interface; The network service module is used to provide the updated knowledge base to the client.

13. An electronic device, characterized in that: include: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

14. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable the computer to execute the method according to any one of claims 1 to 10.

Images