Intelligent recommendation method and system for personalized learning resources based on knowledge graph

Abstract

The application discloses a knowledge graph-based personalized learning resource intelligent recommendation method and system, and relates to the technical field of intelligent recommendation, and the method comprises the following steps: acquiring a knowledge point set corresponding to a classroom teaching target; mapping the knowledge point set based on a knowledge graph to generate a historical learning data set; calculating the historical mastery depth value of a student at a knowledge node, marking a knowledge node lower than a preset mastery degree threshold value as a weak knowledge point to form a weak knowledge point set; planning a personalized learning scheme, determining the learning order and the expected time of the knowledge node, traversing and combining the weak knowledge point set to determine the total number of weak knowledge points; calculating the matching value of the student and the learning scheme; constructing a recommendation model to calculate the recommendation score of the learning scheme; traversing the learning scheme to obtain a recommendation score set; selecting the highest score scheme for pushing; and the application realizes personalized learning resource intelligent recommendation.

Description

Technical Field

[0001] This invention relates to the field of intelligent recommendation technology, specifically to a method and system for intelligent recommendation of personalized learning resources based on knowledge graphs. Background Technology

[0002] With the widespread adoption of educational informatization and personalized teaching models, classroom teaching is integrating with digital learning. Various learning resources are showing a trend of massive growth and multimodal presentation. Personalized learning resource recommendations serve as the core support for teaching and individualized instruction. Their suitability, accuracy, and scientific nature directly affect students' knowledge acquisition, learning efficiency, and the achievement of classroom teaching goals.

[0003] However, traditional learning resource recommendation methods often face the following problems when dealing with scenarios such as massive resource screening, knowledge point association matching, and learning situation adaptation: First, the accuracy of recommendations is insufficient, which can easily lead to information overload and learning confusion. Traditional recommendations rely heavily on collaborative filtering and simple keyword matching, without combining the subject knowledge system and the internal logic of knowledge points. The recommended resources are out of touch with students' actual learning needs and knowledge mastery status. Second, they lack the ability to connect semantic knowledge and intelligently diagnose weaknesses. Traditional recommendations do not construct subject knowledge graphs and cannot explore the prerequisites, connections, and inclusion relationships between knowledge points. They are based only on superficial analysis of learning behavior, making it difficult to identify students' weak knowledge points. The recommended content lacks relevance and cannot effectively fill knowledge gaps. In addition, the recommendation model is static and rigid, lacking personalized learning planning and dynamic adaptation capabilities. Traditional methods mostly push single resources without combining students' historical learning data and learning time to plan personalized learning plans. They also do not quantitatively evaluate the matching degree between learning plans and students' learning situations, and cannot adapt to the different learning paces and ability differences of different students. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for intelligent recommendation of personalized learning resources based on knowledge graphs, so as to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a personalized learning resource intelligent recommendation method based on knowledge graphs, the method comprising the following steps: Obtain the set of knowledge points corresponding to the classroom teaching objectives; Based on the knowledge graph mapping of knowledge point sets, a historical learning data set is generated; In the knowledge graph, the nodes are knowledge nodes and correspond to teaching knowledge points, and the edges correspond to the relationships between knowledge points; Calculate the student's historical mastery depth at each knowledge node, and mark knowledge nodes that are below the preset mastery threshold as weak knowledge points, forming a set of weak knowledge points; Plan a personalized learning plan, determine the learning order and estimated time for knowledge nodes, traverse and combine with the set of weak knowledge points to make a judgment, and count the total number of weak knowledge points. Calculate the match value between students and learning plans; Based on the matching value and the estimated resource time for each knowledge node in the learning scheme, a recommendation model is constructed to calculate the recommendation score of the learning scheme; the learning schemes are traversed to obtain the set of recommendation scores; the scheme with the highest score is selected for push, and historical data is updated or the next scheme is recommended based on the user's completion status.

[0006] Obtain the set of knowledge points corresponding to the classroom teaching objectives; based on the knowledge point set mapped by the knowledge graph, generate a set of historical learning data. Specific steps include: Obtain the classroom teaching objectives and extract the set of knowledge points covered by the classroom teaching objectives; The knowledge point set is mapped based on the knowledge graph, where the nodes in the knowledge graph are called knowledge nodes, the knowledge nodes correspond to teaching knowledge points, and the edges in the knowledge graph correspond to the prerequisite relationships, association relationships, or inclusion relationships between knowledge points. Specifically, each teaching knowledge point covered by the classroom teaching objectives is mapped to a knowledge node in a knowledge graph, so that each knowledge node uniquely corresponds to a teaching knowledge point. Based on the inherent logical relationships between knowledge points, edges are constructed between the corresponding knowledge nodes. These edges represent prerequisite relationships, association relationships, or inclusion relationships between knowledge points. A prerequisite relationship indicates that learning one knowledge point depends on mastering another; an association relationship indicates that multiple knowledge points have parallel or mutually supporting connections; and an inclusion relationship indicates that a knowledge point is composed of several sub-knowledge points. Through the above mapping, the set of knowledge points is transformed into a structured knowledge graph subgraph composed of knowledge nodes and edges. Generate a historical learning dataset {H1, H2, ..., H} k ,...,H K}, where H k This represents the historical learning data of the kth student, where K represents the total number of students in the class. H is a random selection of a student's historical learning data. a The historical learning data includes the student's historical answer accuracy rate (r) at each knowledge node in the knowledge graph. an and the duration of historical learning t an , where r an t represents the accuracy rate of student a's historical answers at knowledge node n. an This represents the duration of student a's historical learning at knowledge node n, where n = 1, 2, ..., N, and N represents the total number of knowledge points covered by the classroom teaching objectives, and a ∈ {1, 2, ..., K}.

[0007] Calculate the student's historical mastery depth at each knowledge node, and mark knowledge nodes below a preset mastery threshold as weak knowledge points, forming a set of weak knowledge points. Specific steps include: The historical mastery depth of student a at knowledge node n is calculated, defined as follows: d an =[r an +(r an / t an )]×w n ; where d an w represents the historical depth of knowledge mastery of student a at knowledge node n. n This represents the weight coefficient of the pre-defined knowledge node n in the classroom teaching objectives; By traversing through the students' historical learning data, the depth of their historical mastery at each knowledge node is obtained. Knowledge points whose depth of historical understanding is less than a preset threshold are marked, forming a set of students' weak historical knowledge points, designated as B. a B a ={b a1 ,b a2 ,...,b an' ,...,b aN'}, b an' This indicates that student a's historical mastery depth at knowledge node n' is lower than the preset mastery threshold, where N' represents the total number of knowledge points that the student lacks mastery of among the knowledge points covered by the classroom teaching objectives, and B a This represents the set of historical knowledge gaps for student a. The preset mastery threshold is determined using percentiles. For any knowledge node n covered by the classroom teaching objectives, the set of historical mastery depth values ​​{d} of all K students in the class at that knowledge node is obtained. 1n ,d 2n ,...,d kn ,...,d Kn}, calculate the Pth percentile of the set as the preset mastery threshold for that knowledge node, where P represents the percentile coefficient; when student a has a historical mastery depth value d for knowledge node n. an If the level of mastery is less than the preset threshold, the knowledge node will be marked as a weak knowledge point.

[0008] Plan a personalized learning program, determine the learning order and estimated time for each knowledge node, iterate through and combine this information with the set of weak knowledge points to make a judgment, and count the total number of weak knowledge points. Specific steps include: Based on the students' current knowledge base and the set of knowledge points covered by the classroom teaching objectives, a learning plan is planned to generate a personalized learning plan. One of these plans is arbitrarily selected and denoted as S.c According to learning scheme S c Learning schemes S were obtained respectively. c The knowledge node corresponding to the i-th learning sequence is L(ci) and the learning scheme is S. c The estimated time T(ci) required for learning the recommended resources at the i-th knowledge node of the plan is given by the time taken, where T(ci) represents the time required for the student to learn the recommended resources at the i-th knowledge node of the plan. c The estimated time required to learn recommended resources at the i-th knowledge node in the plan, L(ci) represents the time required according to the learning scheme S. c The knowledge node corresponding to the i-th learning sequence; The personalized learning plan is designed as follows: Based on the set of knowledge nodes that students have already mastered and the set of knowledge nodes covered by the classroom teaching objectives, and combined with the prerequisite, association, and inclusion relationships between knowledge nodes in the knowledge graph, multiple candidate learning plans are generated. Each learning plan uses knowledge nodes as the basic unit, and the learning order of knowledge nodes is determined according to the prerequisite relationship constraints in the knowledge graph, ensuring that any knowledge node appears after its dependent preceding knowledge node. The learning plans are divided into several types according to different focuses, including complete coverage plans, weak point focus plans, and key node reinforcement plans. The complete coverage plan includes all knowledge nodes covered by the teaching objectives, the weak point focus plan only includes knowledge nodes in the set of students' weak knowledge points and necessary dependent nodes, and the key node reinforcement plan supplements the weak points with key nodes that connect the preceding and following knowledge. After each learning plan is generated, the arrangement order of the knowledge nodes is recorded. The estimated time is determined based on a combination of the recommended resource attributes corresponding to the knowledge node and the individual learning characteristics of the student, specifically as follows: For any knowledge node in the learning plan, obtain the set of recommended resources associated with that knowledge node, including teaching videos, exercises, and explanations of knowledge points, and extract the preset standard time for each resource. At the same time, obtain the actual time the student has spent completing various resources in the past learning process, and calculate the student's learning speed coefficient for each type of resource. Specifically, for each type of resource, divide the sum of the actual time taken by the student to complete all records of that type of resource in the past by the sum of the corresponding preset standard time to obtain the learning speed coefficient for that type of resource. Based on the types of recommended resources associated with the knowledge node, the comprehensive learning speed coefficient corresponding to the knowledge node is determined. Specifically, if the knowledge node is associated with only a single type of resource, the learning speed coefficient of that type of resource is used as the comprehensive learning speed coefficient; if the knowledge node is associated with multiple types of resources, the learning speed coefficients of each type of resource are weighted and averaged according to the standard time ratio of each type of resource to obtain the comprehensive learning speed coefficient. Multiply the sum of the preset standard times for all recommended resources under the knowledge node by the comprehensive learning speed coefficient to obtain the student's estimated time on the knowledge node; When a student completes the (i-1)th knowledge node in the current learning plan and begins learning the ith knowledge node, the student's historical mastery depth value d at that knowledge node is calculated based on their historical learning data. a,L(ci) , where d a,L(ci) Represents the historical mastery depth value of student a at knowledge node L(ci), and queries the set B of the student's historical weak knowledge points. a If L(ci)∈B a Then mark that knowledge node as a weak knowledge point for student a; for learning plan S c Iterate through all knowledge nodes in the learning plan and count the number of nodes marked as weak knowledge points, Z, where Z represents the cumulative number of weak knowledge points of the student in the learning plan.

[0009] The specific steps for calculating the match value between students and learning plans include: The matching value between students and learning plans is calculated, defined as follows: E c =1-Z / I; where E c Represents students and learning plans S c The matching value between them, where I represents the total number of knowledge nodes included in the learning scheme.

[0010] Based on the matching value and the estimated resource usage time for each knowledge node in the learning plan, a recommendation model is constructed to calculate the recommendation score of the learning plan; the learning plans are traversed to obtain the set of recommendation scores; the plan with the highest score is selected for push, and historical data is updated or the next plan is recommended based on the user's completion status. Specific steps include: Construct a learning resource recommendation model, defined as follows: Y c =E c / Σ i=1 I T(ci); where Y c Represents learning scheme S c Recommended score; The generated learning schemes are iterated through to obtain the recommended score set of all learning schemes; Based on the recommended scores of the learning schemes, the learning scheme with the highest recommended score is selected as the push result. A sequence of learning resources is determined and resources are pushed to the user. If the user completes all the learning resources in the pushed learning scheme, the learning progress is recorded and the user's historical learning data is updated. If the user does not complete the learning or actively requests to change the scheme, the next learning scheme with the highest score is recommended until the user completes the learning or all learning schemes have been tried.

[0011] A knowledge graph-based intelligent recommendation system for personalized learning resources, comprising: Data acquisition module, knowledge mapping module, solution planning module, and recommendation and push module; The output of the data acquisition module is connected to the input of the knowledge mapping module; the output of the knowledge mapping module is connected to the output of the scheme planning module; and the output of the scheme planning module is connected to the input of the recommendation push module.

[0012] The data acquisition module includes a target acquisition unit and a history acquisition unit; The target acquisition unit acquires the classroom teaching objectives and extracts the set of knowledge points covered. The history acquisition unit is used to acquire students' historical learning data, including the correct answer rate and learning time for each knowledge node; The knowledge mapping module includes a graph construction unit, a deep computing unit, and a weak point identification unit; The graph construction unit is used to map knowledge points to knowledge nodes based on the knowledge graph and establish prerequisite, association, and inclusion relationships between nodes. The depth calculation unit is used to calculate the student's historical mastery depth value at each knowledge node; The weakness identification unit is used to determine a preset mastery threshold, mark weak knowledge points, and form a weakness set.

[0013] The scheme planning module includes a scheme generation unit, a time estimation unit, and a matching calculation unit; The scheme generation unit is used to generate personalized learning schemes based on students' current mastery and teaching objectives; The time estimation unit estimates the estimated time required for learning resources at each knowledge node in the learning scheme; The matching calculation unit is used to determine the number of weak nodes in the learning scheme by combining the weak set, and to calculate the matching value between the student and the learning scheme. The recommendation push module includes a score calculation unit, a scheme filtering unit, and a feedback update unit; The scoring calculation unit is used to build a recommendation model based on the matching value and resource usage time, and to calculate the recommendation score of each learning scheme. The scheme filtering unit is used to traverse all learning schemes and select the scheme with the highest score as the push result. The feedback update unit is used to update historical data based on the user's completion status, or to continue recommending the next solution.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. By constructing a subject knowledge point association system based on knowledge graphs, and combining students' historical answer accuracy and learning time with multi-dimensional learning data, the depth of knowledge mastery is calculated, and students' weak knowledge points are identified. 2. By combining the learning sequence, estimated time, and matching degree of learning situation, the recommended score of the scheme is calculated, the optimal learning scheme is selected, and historical data is updated according to the learning progress. Unlike the static and single resource push in the existing technology, the present invention can automatically generate a suitable personalized learning scheme and push the best one according to the differences in students' learning situation and learning status. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the intelligent recommendation method for personalized learning resources based on knowledge graphs according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] like Figure 1 As shown, this invention provides a technical solution: a personalized learning resource intelligent recommendation method based on knowledge graphs, which includes the following steps: Obtain the set of knowledge points corresponding to the classroom teaching objectives; Based on the knowledge graph mapping of knowledge point sets, a historical learning data set is generated; In the knowledge graph, the nodes are knowledge nodes and correspond to teaching knowledge points, and the edges correspond to the relationships between knowledge points; Calculate the student's historical mastery depth at each knowledge node, and mark knowledge nodes that are below the preset mastery threshold as weak knowledge points, forming a set of weak knowledge points; Plan a personalized learning plan, determine the learning order and estimated time for knowledge nodes, traverse and combine with the set of weak knowledge points to make a judgment, and count the total number of weak knowledge points. Calculate the match value between students and learning plans; Based on the matching value and the estimated resource time for each knowledge node in the learning scheme, a recommendation model is constructed to calculate the recommendation score of the learning scheme; the learning schemes are traversed to obtain the set of recommendation scores; the scheme with the highest score is selected for push, and historical data is updated or the next scheme is recommended based on the user's completion status.

[0018] Obtain the set of knowledge points corresponding to the classroom teaching objectives; based on the knowledge point set mapped by the knowledge graph, generate a set of historical learning data. Specific steps include: Obtain the classroom teaching objectives and extract the set of knowledge points covered by the classroom teaching objectives; The knowledge point set is mapped based on the knowledge graph, where the nodes in the knowledge graph are called knowledge nodes, the knowledge nodes correspond to teaching knowledge points, and the edges in the knowledge graph correspond to the prerequisite relationships, association relationships, or inclusion relationships between knowledge points. Specifically, each teaching knowledge point covered by the classroom teaching objectives is mapped to a knowledge node in a knowledge graph, so that each knowledge node uniquely corresponds to a teaching knowledge point. Based on the inherent logical relationships between knowledge points, edges are constructed between the corresponding knowledge nodes. These edges represent prerequisite relationships, association relationships, or inclusion relationships between knowledge points. A prerequisite relationship indicates that learning one knowledge point depends on mastering another; an association relationship indicates that multiple knowledge points have parallel or mutually supporting connections; and an inclusion relationship indicates that a knowledge point is composed of several sub-knowledge points. Through the above mapping, the set of knowledge points is transformed into a structured knowledge graph subgraph composed of knowledge nodes and edges. Generate a historical learning dataset {H1, H2, ..., H} k ,...,H K}, where H k This represents the historical learning data of the kth student, where K represents the total number of students in the class. H is a random selection of a student's historical learning data. a The historical learning data includes the student's historical answer accuracy rate (r) at each knowledge node in the knowledge graph. an and the duration of historical learning t an , where r an t represents the accuracy rate of student a's historical answers at knowledge node n. an This represents the duration of student a's historical learning at knowledge node n, where n = 1, 2, ..., N, and N represents the total number of knowledge points covered by the classroom teaching objectives, and a ∈ {1, 2, ..., K}.

[0019] Calculate the student's historical mastery depth at each knowledge node, and mark knowledge nodes below a preset mastery threshold as weak knowledge points, forming a set of weak knowledge points. Specific steps include: The historical mastery depth of student a at knowledge node n is calculated, defined as follows: d an =[r an +(r an / t an )]×w n ; where d an w represents the historical depth of knowledge mastery of student a at knowledge node n. n This represents the weight coefficient of the pre-defined knowledge node n in the classroom teaching objectives; By traversing through the students' historical learning data, the depth of their historical mastery at each knowledge node is obtained. Knowledge points whose depth of historical understanding is less than a preset threshold are marked, forming a set of students' weak historical knowledge points, designated as B. a B a ={b a1 ,b a2 ,...,b an' ,...,b aN'}, b an' This indicates that student a's historical mastery depth at knowledge node n' is lower than the preset mastery threshold, where N' represents the total number of knowledge points that the student lacks mastery of among the knowledge points covered by the classroom teaching objectives, and B a This represents the set of historical knowledge gaps for student a. The preset mastery threshold is determined using percentiles. For any knowledge node n covered by the classroom teaching objectives, the set of historical mastery depth values ​​{d} of all K students in the class at that knowledge node is obtained. 1n ,d 2n ,...,d kn ,...,d Kn}, calculate the Pth percentile of the set as the preset mastery threshold for that knowledge node, where P represents the percentile coefficient; when student a has a historical mastery depth value d for knowledge node n. an If the level of mastery is less than the preset threshold, the knowledge node will be marked as a weak knowledge point.

[0020] Plan a personalized learning program, determine the learning order and estimated time for each knowledge node, iterate through and combine this information with the set of weak knowledge points to make a judgment, and count the total number of weak knowledge points. Specific steps include: Based on the students' current knowledge base and the set of knowledge points covered by the classroom teaching objectives, a learning plan is planned to generate a personalized learning plan. One of these plans is arbitrarily selected and denoted as S. c According to learning scheme S c Learning schemes S were obtained respectively. c The knowledge node corresponding to the i-th learning sequence is L(ci) and the learning scheme is S. c The estimated time T(ci) required for learning the recommended resources at the i-th knowledge node of the plan is given by the time taken, where T(ci) represents the time required for the student to learn the recommended resources at the i-th knowledge node of the plan. c The estimated time required to learn recommended resources at the i-th knowledge node in the plan, L(ci) represents the time required according to the learning scheme S. c The knowledge node corresponding to the i-th learning sequence; The personalized learning plan is designed as follows: Based on the set of knowledge nodes that students have already mastered and the set of knowledge nodes covered by the classroom teaching objectives, and combined with the prerequisite, association, and inclusion relationships between knowledge nodes in the knowledge graph, multiple candidate learning plans are generated. Each learning plan uses knowledge nodes as the basic unit, and the learning order of knowledge nodes is determined according to the prerequisite relationship constraints in the knowledge graph, ensuring that any knowledge node appears after its dependent preceding knowledge node. The learning plans are divided into several types according to different focuses, including complete coverage plans, weak point focus plans, and key node reinforcement plans. The complete coverage plan includes all knowledge nodes covered by the teaching objectives, the weak point focus plan only includes knowledge nodes in the set of students' weak knowledge points and necessary dependent nodes, and the key node reinforcement plan supplements the weak points with key nodes that connect the preceding and following knowledge. After each learning plan is generated, the arrangement order of the knowledge nodes is recorded. The estimated time is determined based on a combination of the recommended resource attributes corresponding to the knowledge node and the individual learning characteristics of the student, specifically as follows: For any knowledge node in the learning plan, obtain the set of recommended resources associated with that knowledge node, including teaching videos, exercises, and explanations of knowledge points, and extract the preset standard time for each resource. At the same time, obtain the actual time the student has spent completing various resources in the past learning process, and calculate the student's learning speed coefficient for each type of resource. Specifically, for each type of resource, divide the sum of the actual time taken by the student to complete all records of that type of resource in the past by the sum of the corresponding preset standard time to obtain the learning speed coefficient for that type of resource. Based on the types of recommended resources associated with the knowledge node, the comprehensive learning speed coefficient corresponding to the knowledge node is determined. Specifically, if the knowledge node is associated with only a single type of resource, the learning speed coefficient of that type of resource is used as the comprehensive learning speed coefficient; if the knowledge node is associated with multiple types of resources, the learning speed coefficients of each type of resource are weighted and averaged according to the standard time ratio of each type of resource to obtain the comprehensive learning speed coefficient. Multiply the sum of the preset standard times for all recommended resources under the knowledge node by the comprehensive learning speed coefficient to obtain the student's estimated time on the knowledge node; When a student completes the (i-1)th knowledge node in the current learning plan and begins learning the ith knowledge node, the student's historical mastery depth value d at that knowledge node is calculated based on their historical learning data. a,L(ci)' , where d a,L(ci) Represents the historical mastery depth value of student a at knowledge node L(ci), and queries the set B of the student's historical weak knowledge points. a If L(ci)∈B a Then mark that knowledge node as a weak knowledge point for student a; for learning plan Sc Iterate through all knowledge nodes in the learning plan and count the number of nodes marked as weak knowledge points, Z, where Z represents the cumulative number of weak knowledge points of the student in the learning plan.

[0021] The specific steps for calculating the match value between students and learning plans include: The matching value between students and learning plans is calculated, defined as follows: E c =1-Z / I; where E c Represents students and learning plans S c The matching value between them, where I represents the total number of knowledge nodes included in the learning scheme.

[0022] Based on the matching value and the estimated resource usage time for each knowledge node in the learning plan, a recommendation model is constructed to calculate the recommendation score of the learning plan; the learning plans are traversed to obtain the set of recommendation scores; the plan with the highest score is selected for push, and historical data is updated or the next plan is recommended based on the user's completion status. Specific steps include: Construct a learning resource recommendation model, defined as follows: Y c =E c / Σ i=1 I T(ci); where Y c Represents learning scheme S c Recommended score; The generated learning schemes are iterated through to obtain the recommended score set of all learning schemes; Based on the recommended scores of the learning schemes, the learning scheme with the highest recommended score is selected as the push result. A sequence of learning resources is determined and resources are pushed to the user. If the user completes all the learning resources in the pushed learning scheme, the learning progress is recorded and the user's historical learning data is updated. If the user does not complete the learning or actively requests to change the scheme, the next learning scheme with the highest score is recommended until the user completes the learning or all learning schemes have been tried.

[0023] In Example 1: The teacher sets the classroom teaching objectives for this lesson, which cover five teaching knowledge points, namely knowledge point one, knowledge point two, knowledge point three, knowledge point four and knowledge point five; The teaching objective is obtained, and the five knowledge points mentioned above are mapped based on a pre-constructed knowledge graph. In this knowledge graph, each node is called a knowledge node, and each knowledge node corresponds to a specific teaching knowledge point. The edges in the knowledge graph are used to represent the relationships between knowledge points: for example, there is a prerequisite relationship between knowledge node one and knowledge node two, that is, students must master node one before they can learn node two; there is a parallel relationship between knowledge node two and knowledge node three; knowledge node five has a comprehensive application relationship with the previous four knowledge points, and students need to master the prerequisite knowledge points before they can learn it effectively. Subsequently, the historical learning data of all students were obtained from the teaching resource platform. Taking student A as an example, their historical learning data included the accuracy rate of their historical answers and the duration of their historical learning sessions on the five knowledge nodes mentioned above. Based on the historical data mentioned above, the historical mastery depth value of student A at each knowledge node was calculated. Combined with the overall level of the class, the percentile method was used to determine the preset mastery threshold for each knowledge node. By comparison, student A was marked as having weak knowledge points in knowledge nodes two, three, and five, forming a set of weak knowledge points for him / her. After completing the learning analysis, we enter the learning plan planning stage. Based on the prerequisite relationships between knowledge nodes in the knowledge graph, combined with the knowledge nodes that student A has already mastered and the classroom teaching objectives, three candidate learning plans are generated. The first learning plan follows the order of prerequisite relationships in the knowledge graph, learning all five knowledge nodes in sequence. Special time is allocated to strengthen student A's three weak knowledge points, while the knowledge nodes that have already been mastered are quickly reviewed. The second learning plan only targets student A's three weak knowledge points, skipping the knowledge nodes that have already been mastered. However, during the path planning, it was detected that since knowledge node five depends on the mastery of the preceding knowledge nodes, directly learning this node may be difficult. Therefore, a brief review of the preceding knowledge nodes was added to this plan. The third learning plan, while strengthening weak knowledge points, focuses on supplementing knowledge node four, because this node is the key link connecting the previous solution method and the subsequent comprehensive application. Strengthening this node will help student A better master the comprehensive application ability. For each learning plan, a matching value is calculated, which reflects the proportion of weak knowledge points in the learning plan. Then, combined with the estimated time required to learn the recommended resources at each knowledge node, a recommendation score is calculated for each learning plan. The recommendation score takes into account the degree of matching between the plan and the student's knowledge level as well as the time cost required to complete the plan. After comparison, although the first learning plan took the longest time in total, it had the highest matching degree and the time investment was within an acceptable range. Therefore, it received the highest recommendation score and was used as the push result to push the corresponding learning resource sequence to student A. Once student A begins learning, their learning progress is tracked in real time. If student A completes all learning resources, their learning progress is recorded and their historical learning data is updated for subsequent learning analysis. If student A encounters difficulties during the learning process and actively requests a change of plan, the second-highest-scoring learning plan is recommended until student A completes the learning or all plans have been tried.

[0024] A knowledge graph-based intelligent recommendation system for personalized learning resources, comprising: Data acquisition module, knowledge mapping module, solution planning module, and recommendation and push module; The output of the data acquisition module is connected to the input of the knowledge mapping module; the output of the knowledge mapping module is connected to the output of the scheme planning module; and the output of the scheme planning module is connected to the input of the recommendation push module.

[0025] The data acquisition module includes a target acquisition unit and a history acquisition unit; The target acquisition unit acquires the classroom teaching objectives and extracts the set of knowledge points covered. The history acquisition unit is used to acquire students' historical learning data, including the correct answer rate and learning time for each knowledge node; The knowledge mapping module includes a graph construction unit, a deep computing unit, and a weak point identification unit; The graph construction unit is used to map knowledge points to knowledge nodes based on the knowledge graph and establish prerequisite, association, and inclusion relationships between nodes. The depth calculation unit is used to calculate the student's historical mastery depth value at each knowledge node; The weakness identification unit is used to determine a preset mastery threshold, mark weak knowledge points, and form a weakness set.

[0026] The scheme planning module includes a scheme generation unit, a time estimation unit, and a matching calculation unit; The scheme generation unit is used to generate personalized learning schemes based on students' current mastery and teaching objectives; The time estimation unit estimates the estimated time required for learning resources at each knowledge node in the learning scheme; The matching calculation unit is used to determine the number of weak nodes in the learning scheme by combining the weak set, and to calculate the matching value between the student and the learning scheme. The recommendation push module includes a score calculation unit, a scheme filtering unit, and a feedback update unit; The scoring calculation unit is used to build a recommendation model based on the matching value and resource usage time, and to calculate the recommendation score of each learning scheme. The scheme filtering unit is used to traverse all learning schemes and select the scheme with the highest score as the push result. The feedback update unit is used to update historical data based on the user's completion status, or to continue recommending the next solution.

[0027] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A personalized learning resource intelligent recommendation method based on knowledge graphs, characterized by: The method includes the following steps: Obtain the set of knowledge points corresponding to the classroom teaching objectives; Based on the knowledge graph mapping of knowledge point sets, a historical learning data set is generated; In the knowledge graph, the nodes are knowledge nodes and correspond to teaching knowledge points, and the edges correspond to the relationships between knowledge points; Calculate the student's historical mastery depth at each knowledge node, and mark knowledge nodes that are below the preset mastery threshold as weak knowledge points, forming a set of weak knowledge points; Plan a personalized learning plan, determine the learning order and estimated time for knowledge nodes, traverse and combine with the set of weak knowledge points to make a judgment, and count the total number of weak knowledge points. Calculate the match value between students and learning plans; Based on the matching value and the estimated resource time for each knowledge node in the learning scheme, a recommendation model is constructed to calculate the recommendation score of the learning scheme; the learning schemes are traversed to obtain the set of recommendation scores; the scheme with the highest score is selected for push, and historical data is updated or the next scheme is recommended based on the user's completion status.

2. The intelligent recommendation method for personalized learning resources based on knowledge graphs according to claim 1, characterized in that: Obtain the set of knowledge points corresponding to the classroom teaching objectives; Based on the knowledge graph mapping of knowledge point sets, a historical learning data set is generated. The specific steps include: Obtain the classroom teaching objectives and extract the set of knowledge points covered by the classroom teaching objectives; The knowledge point set is mapped based on the knowledge graph, where the nodes in the knowledge graph are called knowledge nodes, the knowledge nodes correspond to teaching knowledge points, and the edges in the knowledge graph correspond to the prerequisite relationships, association relationships, or inclusion relationships between knowledge points. Generate a historical learning dataset {H1, H2, ..., H} K }, where H1, H2, H K These represent the historical learning data of the 1st, 2nd, and Kth students, respectively, where K represents the total number of students in the class. H is a random selection of a student's historical learning data. a The historical learning data includes the student's historical answer accuracy rate (r) at each knowledge node in the knowledge graph. an and the duration of historical learning t an , where r an t represents the historical answer accuracy rate of student a at knowledge node n. an This represents the duration of student a's historical learning at knowledge node n, where n = 1, 2, ..., N, and N represents the total number of knowledge points covered by the classroom teaching objectives, and a ∈ {1, 2, ..., K}.

3. The intelligent recommendation method for personalized learning resources based on knowledge graphs according to claim 2, characterized in that: Calculate the student's historical mastery depth at each knowledge node, and mark knowledge nodes below a preset mastery threshold as weak knowledge points, forming a set of weak knowledge points. Specific steps include: The historical mastery depth of student a at knowledge node n is calculated, defined as follows: d an =[r an +(r an / t an )]×w n ; where d an w represents the historical depth of knowledge mastery of student a at knowledge node n. n This represents the weight coefficient of the pre-defined knowledge node n in the classroom teaching objectives; By traversing through the students' historical learning data, the depth of their historical mastery at each knowledge node is obtained. Knowledge points whose depth of historical understanding is less than a preset threshold are marked, forming a set of students' weak historical knowledge points, designated as B. a B a This represents the set of historical knowledge gaps for student a. The preset mastery threshold is determined using a percentile approach. For any knowledge node n covered by the classroom teaching objectives, the set of historical mastery depth values ​​for all K students in the class at that knowledge node is obtained, and the Pth percentile of this set is calculated as the preset mastery threshold for that knowledge node, where P represents the percentile coefficient. When student a's historical mastery depth value d at knowledge node n... an If the level of mastery is less than the preset threshold, the knowledge node will be marked as a weak knowledge point.

4. The intelligent recommendation method for personalized learning resources based on knowledge graphs according to claim 3, characterized in that: Plan a personalized learning program, determine the learning order and estimated time for each knowledge node, iterate through and combine this information with the set of weak knowledge points to make a judgment, and count the total number of weak knowledge points. Specific steps include: Based on the students' current knowledge base and the set of knowledge points covered by the classroom teaching objectives, a learning plan is planned to generate a personalized learning plan. One of these plans is arbitrarily selected and denoted as S. c According to learning scheme S c Learning schemes S were obtained respectively. c The knowledge node corresponding to the i-th learning sequence is L(ci) and the learning scheme is S. c The estimated time T(ci) required for learning the recommended resources at the i-th knowledge node in the plan is given by the time taken, where T(ci) represents the time required for the student to learn the recommended resources at the i-th knowledge node in the plan S. c The estimated time required to learn recommended resources at the i-th knowledge node in the plan, L(ci) represents the time required according to the learning scheme S. c The knowledge node corresponding to the i-th learning sequence; When a student completes the (i-1)th knowledge node in the current learning plan and begins learning the ith knowledge node, the student's historical mastery depth value d at that knowledge node is calculated based on their historical learning data. a,L(ci) , where d a,L(ci) Represents the historical mastery depth value of student a at knowledge node L(ci), and queries the set B of the student's historical weak knowledge points. a If L(ci)∈B a Then mark that knowledge node as a weak knowledge point for student a; for learning plan S c Iterate through all knowledge nodes in the learning plan and count the number of nodes marked as weak knowledge points, Z, where Z represents the cumulative number of weak knowledge points of the student in the learning plan.

5. The intelligent recommendation method for personalized learning resources based on knowledge graphs according to claim 4, characterized in that: The specific steps for calculating the match value between students and learning plans include: The matching value between students and learning plans is calculated, defined as follows: E c =1-Z / I; where E c Represents students and learning plans S c The matching value between them, where I represents the total number of knowledge nodes included in the learning scheme.

6. The intelligent recommendation method for personalized learning resources based on knowledge graphs according to claim 5, characterized in that: Based on the matching value and the estimated resource usage time for each knowledge node in the learning scheme, a recommendation model is constructed to calculate the recommendation score of the learning scheme. The learning schemes are iterated to obtain a set of recommended scores; the scheme with the highest score is selected for recommendation, and historical data is updated or the next scheme is recommended based on the user's completion status. Specific steps include: Construct a learning resource recommendation model, defined as follows: Y c =E c / Σ i=1 I T(ci); where Y c Represents learning scheme S c Recommended score; The generated learning schemes are traversed to obtain the recommended score set of all learning schemes; Based on the recommended scores of the learning schemes, the learning scheme with the highest recommended score is selected as the push result. A sequence of learning resources is determined and resources are pushed to the user. If the user completes all the learning resources in the pushed learning scheme, the learning progress is recorded and the user's historical learning data is updated. If the user does not complete the learning or actively requests to change the scheme, the next learning scheme with the highest score is recommended until the user completes the learning or all learning schemes have been tried.

7. A knowledge graph-based intelligent recommendation system for personalized learning resources, applied to the knowledge graph-based intelligent recommendation method for personalized learning resources as described in any one of claims 1-6, characterized in that: The system includes: Data acquisition module, knowledge mapping module, solution planning module, and recommendation and push module; The output of the data acquisition module is connected to the input of the knowledge mapping module; the output of the knowledge mapping module is connected to the output of the scheme planning module; and the output of the scheme planning module is connected to the input of the recommendation push module.

8. The knowledge graph-based intelligent recommendation system for personalized learning resources according to claim 7, characterized in that: The data acquisition module includes a target acquisition unit and a history acquisition unit; The target acquisition unit acquires the classroom teaching objectives and extracts the set of knowledge points covered. The history acquisition unit is used to acquire students' historical learning data, including the correct answer rate and learning time for each knowledge node; The knowledge mapping module includes a graph construction unit, a deep computing unit, and a weak point identification unit; The graph construction unit is used to map knowledge points to knowledge nodes based on the knowledge graph and establish prerequisite, association, and inclusion relationships between nodes. The depth calculation unit is used to calculate the student's historical mastery depth value at each knowledge node; The weakness identification unit is used to determine a preset mastery threshold, mark weak knowledge points, and form a weakness set.

9. The knowledge graph-based intelligent recommendation system for personalized learning resources according to claim 8, characterized in that: The scheme planning module includes a scheme generation unit, a time estimation unit, and a matching calculation unit; The scheme generation unit is used to generate personalized learning schemes based on students' current mastery and teaching objectives; The time estimation unit estimates the estimated time required for learning resources at each knowledge node in the learning scheme; The matching calculation unit is used to determine the number of weak nodes in the learning scheme by combining the weak set, and to calculate the matching value between the student and the learning scheme. The recommendation push module includes a score calculation unit, a scheme filtering unit, and a feedback update unit; The scoring calculation unit is used to build a recommendation model based on the matching value and resource usage time, and to calculate the recommendation score of each learning scheme. The scheme filtering unit is used to traverse all learning schemes and select the scheme with the highest score as the push result. The feedback update unit is used to update historical data based on the user's completion status, or to continue recommending the next solution.

Images