Learning resource intelligent management system and method based on data analysis

Abstract

The invention discloses a learning resource intelligent management system and method based on data analysis, and belongs to the technical field of resource management. The method comprises the following steps: traversing multi-platform courses, extracting knowledge points, constructing a knowledge graph, and summarizing to form a multi-platform knowledge point database; repeated knowledge points are screened, and explanation emphasis directions of all the platforms are analyzed and quantified into specific numerical values. According to historical test data of students, answer expressions are disassembled to deduce knowledge point emphasis directions and quantify the knowledge point emphasis directions into specific numerical values, and courses with the most similar emphasis degrees are recommended after comparison. And if the matching degrees of the multiple courses are consistent, constructing a multivariate quantitative scoring system containing indexes such as teacher qualification and course favorable comments, carrying out weighted calculation on comprehensive scores, and carrying out sorting recommendation. Cross-platform resource integration and accurate personalized learning resource pushing are realized in the whole process, and the learning efficiency and the resource utilization efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of resource management technology, specifically to an intelligent management system and method for learning resources based on data analysis. Background Technology

[0002] With the rapid development of internet technology, online education, with its advantages of breaking through time and space limitations and having wide resource coverage, has become an important development direction in the education field. Various online education platforms have sprung up like mushrooms after rain, covering multiple fields such as academic education, vocational skills training, and certification tutoring, forming a massive reserve of course resources. Against this backdrop, students' demand for personalized learning resources is becoming increasingly prominent, hoping to quickly find course content that suits their knowledge gaps and learning preferences.

[0003] Existing technologies are largely limited to course management within a single platform, lacking the ability to comprehensively cover and integrate course resources across multiple platforms. They cannot systematically identify repetitive knowledge points across different platforms, forcing students to repeatedly search across multiple platforms, resulting in low learning efficiency and a waste of valuable resources. For repetitive knowledge points across platforms, existing technologies do not systematically analyze and quantify the teaching emphasis of each platform, relying solely on superficial information such as course titles and descriptions for recommendations, failing to accurately match students' specific needs for knowledge points. Furthermore, existing technologies do not deeply analyze students' historical test data reflecting their knowledge mastery, making it impossible to accurately deduce students' core focus on each knowledge point, resulting in a lack of targeted needs assessment and difficulty in supporting personalized recommendations. Summary of the Invention

[0004] The purpose of this invention is to provide a data analysis-based intelligent management system and method for learning resources to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: Firstly, this application provides a data analysis-based intelligent management method for learning resources, comprising the following steps: Iterate through all courses on various online education platforms and extract the knowledge points covered in each course; for a single course, sort out the progressive and horizontal relationships between knowledge points and construct the knowledge graph of the course; summarize the knowledge graphs of all courses on all platforms to form a multi-platform knowledge point database; Based on a multi-platform knowledge point database, we screened out repetitive knowledge points that exist on two or more platforms; for each repetitive knowledge point, we analyzed the emphasis of different platforms in explaining it; and we established a quantitative evaluation model to convert the degree of emphasis of each platform on the repetitive knowledge point into specific values. Based on students' historical test data, we analyze their performance on each knowledge point and deduce their emphasis on each knowledge point. Through a quantitative evaluation model, we convert the degree of students' emphasis on each knowledge point into a quantitative value. We then compare the quantitative results of students' emphasis on repetitive knowledge points with those of various platforms and recommend courses with the closest degree of emphasis to students. When multiple courses align with a student's focus, a multi-dimensional quantitative scoring system is constructed, assigning corresponding weights to each indicator and calculating a comprehensive score for each course. Courses are then ranked from highest to lowest based on their comprehensive scores, and the ranked courses are recommended to the corresponding students.

[0006] In conjunction with the first aspect, in the first embodiment of the first aspect of this application, the step of traversing all courses on each online education platform and extracting the knowledge points covered by each course includes: Identify the online education platforms to be covered, define the types and status of courses to be processed on each platform, and determine the traversal boundaries; obtain the basic information of all courses on each platform in batches through platform open interface calls and standardized data crawling; for platforms without open interfaces, use automated scripts to traverse the course list page according to preset rules, locate and obtain the basic information of individual courses one by one; For each course, its core teaching resources are analyzed, redundant information is removed, and text and structured content related to the core teaching are extracted to form a set of teaching content materials for a single course. Based on the teaching content material set, knowledge points with independent teaching significance are identified and separated through semantic analysis, keyword clustering, and teaching logic decomposition. For content with hierarchical relationships, it is first separated into the finest-grained independent knowledge points, and then its level is recorded. The extracted independent knowledge points are initially screened for repetition, and the accuracy and completeness of the knowledge point descriptions are verified. Ambiguous, unclear, or content without actual teaching significance is removed to form a list of independent knowledge points for a single course.

[0007] In conjunction with the first aspect, in the second embodiment of the first aspect of this application, the step of sorting out the progressive and horizontal relationships between knowledge points and constructing a knowledge graph for a single course includes: Each independent knowledge point in a single course is labeled with basic attributes; based on the basic attributes of the knowledge points, the course teaching logic, cognitive learning patterns, and difficulty levels of the knowledge points are extracted, the progressive dependencies between knowledge points are sorted out, the order of learning knowledge points first and later is clarified, the support relationship between basic knowledge points and advanced knowledge points is determined, and a list of progressive relationships is formed; based on the basic attributes of the knowledge points, the knowledge point types, teaching application scenarios, and the logic of the after-class exercises are extracted, the horizontal connections between knowledge points at the same level or across levels are sorted out, different knowledge points for solving the same problem and collaborative knowledge points in the same application scenario are identified, the connection dimensions between each knowledge point are clarified, and a list of horizontal connection relationships is formed. Using independent knowledge points as nodes and the progressive relationships as edges, a structured knowledge graph for a single course is constructed, and the basic attributes and relationship descriptions of each node are marked in the graph.

[0008] In conjunction with the first aspect, in the third embodiment of the first aspect of this application, the step of filtering repetitive knowledge points existing on two or more platforms based on a multi-platform knowledge point database includes: Based on a multi-platform knowledge point database, this method compares the basic attributes of knowledge points to determine the intrinsic equivalence between different expressions. All equivalent non-standard expressions are then reduced to the same standard expression, forming a one-way mapping table from non-standard to standard expressions. This mapping table is used to clean the original dataset. Using the cleaned standardized knowledge points as the primary index key, the same platform ID is counted only once, regardless of how many courses it corresponds to, thus constructing a one-to-one correspondence index structure between standardized knowledge points and platform sets. A threshold for duplicate detection is set; when a knowledge point exists on two or more platforms, it is considered a duplicate knowledge point. All index entries are traversed, and the number of elements in the platform set corresponding to each standardized knowledge point is calculated to obtain the counting result. The counting result for each knowledge point is compared with the preset threshold, and knowledge points whose counting results meet the threshold condition are selected to form a set of duplicate knowledge points.

[0009] In conjunction with the first aspect, in the fourth embodiment of the first aspect of this application, the step of analyzing the emphasis of different platforms in explaining each repetitive knowledge point includes: For each repetitive knowledge point, teaching-related text data corresponding to that knowledge point is extracted from each platform; redundant information is removed from the teaching-related text data, and the text is segmented using the jieba word segmentation tool to form a structured text dataset corresponding to a single repetitive knowledge point on each platform; the structured text datasets from each platform are truncated or padded according to the model requirements and input into the BERT model; the bidirectional Transformer architecture of the BERT model is used to capture the contextual semantic relationships of core words in the text, and the text is transformed into a fixed-dimensional semantic feature vector; The K-Means unsupervised clustering algorithm is used to mine topics from the semantic feature vectors of individual repetitive knowledge points. Specifically, the optimal number of clusters is determined based on the elbow rule; the semantic feature vectors corresponding to each platform are input into the K-Means algorithm, which iteratively calculates and minimizes the sum of squares within each cluster, aggregating semantic feature vectors with high semantic similarity into the same topic cluster, with each topic cluster corresponding to a potential knowledge focus direction; after clustering, keywords are extracted from the text fragments within each topic cluster to generate topic tags. For each topic cluster to which the text data belongs on each platform, the TF-IDF algorithm is used to calculate the weight of each topic cluster. Specifically, taking the platform's text dataset as a unit, the word frequency (TF) of the text fragment corresponding to each topic cluster in the platform's dataset is counted. Combined with the inverse document frequency (IDF) of the keywords of the topic cluster in all platform datasets, the TF-IDF weight value of each topic cluster is calculated. The topic cluster with the highest weight value is selected, and its corresponding topic tag is determined as the platform's focus on the knowledge direction of this repetitive knowledge point.

[0010] In conjunction with the first aspect, in the fifth embodiment of the first aspect of this application, the establishment of a quantitative evaluation model, which converts the degree of emphasis on the repetitive knowledge point by each platform into specific numerical values, includes: By associating repetitive knowledge points with corresponding platforms and key knowledge areas, and breaking down the constituent dimensions of each key knowledge area, a multi-dimensional quantitative evaluation system is established, clarifying the evaluation connotation and value range definition rules for each dimension. For each repetitive knowledge point and its corresponding platform, basic data for each evaluation dimension is extracted from teaching text datasets and course structure information. A normalization mathematical algorithm is used to map the basic data of different dimensions to a unified numerical range, eliminating evaluation bias caused by differences in the units of measurement between different dimensions, and obtaining standardized scores for each dimension. Based on the influence weight of each evaluation dimension on the degree of emphasis, a weight allocation algorithm is used to determine the weight coefficient of each dimension. A weighted aggregation algorithm is then used to calculate the weighted average of the standardized scores of each dimension with their corresponding weight coefficients, resulting in a comprehensive quantitative value of the degree of emphasis on the repetitive knowledge point for a single platform. The comprehensive quantitative values ​​of adjacent repetitive knowledge points within the same platform are logically compared, and the distinguishability of the quantitative results of the same repetitive knowledge point across different platforms is verified to eliminate outliers caused by data extraction bias or unreasonable weight allocation.

[0011] In conjunction with the first aspect, in the sixth embodiment of the first aspect of this application, the step of analyzing students' historical test data to break down their performance on each knowledge point and deduce their emphasis on each knowledge point includes: Collect students' historical test data, clean the data, and remove invalid data; based on the historical test data, calculate the percentage of correct answers for each student across all questions corresponding to that knowledge point, reflecting the overall mastery of the knowledge point; categorize and statistically analyze the incorrect questions for that knowledge point according to preset error types, and identify the core causes of errors; calculate the average answering time for each question corresponding to that knowledge point, and combine this with the time taken based on the difficulty level of the questions to determine the student's understanding and proficiency of the knowledge point; comprehensively form multi-dimensional statistical results of each student's answering performance for each knowledge point; The study systematically analyzes the types of knowledge points that emphasize specific knowledge directions, establishes a system of rules linking answer performance dimensions with the knowledge directions emphasized, and clarifies the emphasis requirements corresponding to different combinations of answer performance. It then matches the statistical results of each student's multi-dimensional answer performance for each knowledge point with the pre-defined system of rules. Finally, it employs a weighted judgment algorithm to rank the matched potential emphasis directions by weight, selecting the direction with the highest weight as the student's emphasis on that knowledge point.

[0012] In conjunction with the first aspect, in the seventh embodiment of the first aspect of this application, the step of comparing the quantitative results of the students' emphasis on repetitive knowledge points with those of each platform one by one, and recommending courses with the closest emphasis to the students, includes: Based on the dimensions of individual students and individual repetitive knowledge points, the student's emphasis on that knowledge point is quantified and compared with the quantified values ​​of the corresponding knowledge points on each platform. The platforms are then sorted by the degree of difference to determine the degree of matching between the students and each platform. For each repetitive knowledge point, the platform courses that are closest to the student's emphasis are selected. The best matching courses for all knowledge points are compiled, and after removing duplicate courses, a personalized recommendation list is formed, thus completing the recommendation output.

[0013] In conjunction with the first aspect, in the eighth embodiment of the first aspect of this application, the step of constructing a multi-dimensional quantitative scoring system when multiple courses align with the student's focus, assigning corresponding weights to each indicator, and calculating the comprehensive score for each course includes: Indicators affecting course quality were selected and incorporated into a multi-dimensional quantitative scoring system, including teacher qualifications, course ratings, and overall course duration. Based on the degree of influence of each indicator on course quality, a weighting allocation logic was used to determine the weight coefficient of each type of indicator, ensuring that the total weight is 1. The original data of the corresponding indicators for each course were extracted and standardized. The standardized data and the corresponding indicator weights were then weighted and aggregated to obtain the comprehensive score for each course.

[0014] Secondly, this application provides a data analysis-based intelligent management system for learning resources, comprising: The knowledge graph construction module includes: a data acquisition unit that traverses all courses on various online education platforms and extracts the knowledge points covered by each course; a single-course knowledge graph construction unit that, for a single course, sorts out the progressive and horizontal relationships between knowledge points and constructs the knowledge graph of the course; and a multi-platform knowledge point database aggregation unit that aggregates the knowledge graphs of all courses on all platforms to form a multi-platform knowledge point database. The module for analyzing and quantifying repetitive knowledge points includes: a repetitive knowledge point screening unit that filters repetitive knowledge points existing on two or more platforms based on a multi-platform knowledge point database; a platform-emphasized knowledge direction analysis unit that analyzes the emphasis of different platforms on each repetitive knowledge point; and a quantification unit that establishes a quantitative evaluation model to convert the degree of emphasis of each platform on the repetitive knowledge point into specific numerical values. The course matching module includes: a student-focused knowledge direction deduction unit that analyzes students' historical test data to break down their performance on each knowledge point and deduce their focus on each knowledge point; a student-focused degree quantification unit that uses a quantitative evaluation model to convert the degree of students' focus on each knowledge point into a quantitative value; and a quantitative result comparison and matching unit that compares the quantitative results of students' focus on repetitive knowledge points with those of various platforms one by one and recommends courses with the closest focus to students. The course recommendation optimization module includes: a course comprehensive score calculation unit that constructs a multi-dimensional quantitative scoring system when multiple courses align with a student's focus, assigns corresponding weights to each indicator, and calculates the comprehensive score for each course; and a course ranking and recommendation unit that ranks courses from highest to lowest comprehensive score and recommends the ranked courses to the corresponding students.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention, by traversing courses across multiple platforms, constructing a knowledge graph, and forming a comprehensive knowledge point database, can systematically identify repetitive knowledge points across different platforms, breaking down resource barriers of a single platform, providing a comprehensive resource foundation for subsequent accurate matching, and improving the utilization rate of high-quality cross-platform resources.

[0016] 2. This invention uses AI algorithms to analyze the emphasis of various platforms on repetitive knowledge points and establishes a quantitative evaluation model to convert the degree of emphasis into specific values. At the same time, it combines students' historical test data to deduce and quantify their emphasis needs. By comparing the quantitative results, it achieves accurate matching between students' needs and platform courses, which greatly improves the targeting of recommendations.

[0017] 3. This invention takes students' historical test data as the core and derives the focus of knowledge points by breaking down the problem-solving performance. Compared with the shallow data mining of existing technologies, it can more accurately capture students' knowledge weaknesses and learning preferences, providing a precise basis for personalized recommendations. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the steps of an intelligent management method for learning resources based on data analysis according to the present invention. Figure 2This is a system architecture diagram of an intelligent management system for learning resources based on data analysis, as proposed in this invention. Detailed Implementation

[0019] 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.

[0020] Example: Figures 1-2 As shown, the present invention provides a technical solution: like Figure 1 As shown, this application provides a data analysis-based intelligent management method for learning resources, including the following steps: Step S100: Traverse all courses on each online education platform and extract the knowledge points covered by each course; for a single course, sort out the progressive and horizontal relationships between knowledge points and construct the knowledge graph of the course; summarize the knowledge graphs of all courses on all platforms to form a multi-platform knowledge point database; Specifically, identify the online education platforms to be covered, define the types and status of courses to be processed on each platform, and determine the traversal boundaries; obtain the basic information of all courses on each platform in batches through platform open interface calls and standardized data crawling; for platforms without open interfaces, use automated scripts to traverse the course list page according to preset rules, locate and obtain the basic information of individual courses one by one; For each course, its core teaching resources are analyzed, redundant information is removed, and text and structured content related to the core teaching are extracted to form a set of teaching content materials for a single course. Based on the teaching content material set, knowledge points with independent teaching significance are identified and separated through semantic analysis, keyword clustering, and teaching logic decomposition. For content with hierarchical relationships, it is first separated into the finest-grained independent knowledge points, and then its level is recorded. The extracted independent knowledge points are initially screened for repetition, and the accuracy and completeness of the knowledge point descriptions are verified. Ambiguous, unclear, or content without actual teaching significance is removed to form a list of independent knowledge points for a single course.

[0021] Furthermore, basic attributes are labeled for each independent knowledge point in a single course; based on the basic attributes of the knowledge points, the course teaching logic, cognitive learning patterns, and difficulty levels of the knowledge points are extracted, the progressive dependencies between knowledge points are sorted out, the order of learning knowledge points first and later is clarified, the support relationship between basic knowledge points and advanced knowledge points is determined, and a list of progressive relationships is formed; based on the basic attributes of the knowledge points, the knowledge point types, teaching application scenarios, and the logic of the after-class exercises are extracted, the horizontal connections between knowledge points at the same level or across levels are sorted out, different knowledge points for solving the same problem and collaborative knowledge points in the same application scenario are identified, the connection dimensions between each knowledge point are clarified, and a list of horizontal connection relationships is formed. Using independent knowledge points as nodes and the progressive relationships as edges, a structured knowledge graph for a single course is constructed, and the basic attributes and relationship descriptions of each node are marked in the graph.

[0022] In one specific embodiment, this embodiment selects three mainstream domestic online education platforms (denoted as Platform A, Platform B, and Platform C) as the traversal objects, defines the course type to be processed as Python programming in the vocational skills category, and limits the course status to those that have been online for at least 30 days (excluding newly launched test courses). The traversal boundary is clearly defined as all compliant courses in this direction within the three platforms. Through platform open interface calls and automated script crawling, basic information of 85 courses were successfully obtained, including 28 Python programming courses on Platform A, 35 on Platform B, and 22 on Platform C. This information includes core data such as course ID, course name, chapter outline, class hour distribution, and courseware resource links. For Platform C, which does not have an interface, the script traversed its course list page to accurately locate and obtain the entry point for each course details, achieving a data acquisition success rate of 98.6%.

[0023] Using the "Python from Zero to Hero" course (Course ID: A-Py-017) on Platform A as a typical sample, knowledge points were extracted. The course outline, 24 lesson scripts, and accompanying PPTs were analyzed to extract core teaching resources. After removing approximately 3200 words of redundant text such as course introductions and advertisements, 12,000 words of text and structured content relevant to the core teaching were extracted, forming a teaching content material set for the course. Based on this material set, semantic analysis and other methods were used to identify 42 knowledge points with independent teaching significance, including "Python variable definition," "list creation and access," and "dictionary key-value pair operations." Content with hierarchical relationships (such as "parameter passing" and "return value setting" under "function advancement") was broken down to the finest granularity. Two knowledge points with duplicate descriptions were removed after initial screening. After verification, a final list of 40 knowledge points was formed, with 100% accuracy in knowledge point description.

[0024] The 40 knowledge points were labeled with basic attributes, including knowledge type (12 concepts, 20 methods, and 8 practical skills) and difficulty level (25 basic, 12 intermediate, and 3 extended). Based on the attribute relationships, 36 sets of progressive dependencies were extracted, clarifying core logic such as "Python variable definition" being a prerequisite for "list creation and access" and "conditional statements" being a prerequisite for "loop statements," forming a list of progressive relationships. 18 sets of horizontal relationships were also identified, such as the collaborative relationship between "list slicing" and "dictionary traversal" in "batch data processing" scenarios, and the substitution relationship between "for loops" and "while loops" for solving "repetitive tasks," forming a list of horizontal relationships.

[0025] Using 40 independent knowledge points as nodes and 36 sets of progressive relationships as edges, a structured knowledge graph for the course was constructed. The graph clearly labels the attributes and relationships of each node, achieving 100% node coverage and complete relationship inclusion. After constructing knowledge graphs for 85 courses using this process, a multi-platform knowledge point database was formed. This database contains 1268 independent knowledge points and 2135 sets of relationships (progressive + horizontal), achieving full integration of knowledge points from Python programming courses across three platforms, with data integrity reaching 97.8%.

[0026] Step S200: Based on a multi-platform knowledge point database, filter out repetitive knowledge points that exist on two or more platforms; for each repetitive knowledge point, analyze the emphasis of different platforms in explaining it; establish a quantitative evaluation model to convert the degree of emphasis of each platform on the repetitive knowledge point into specific values. Specifically, based on a multi-platform knowledge point database, the equivalence of different expressions is determined by comparing the basic attributes of knowledge points. All equivalent non-standard expressions are reduced to the same standard expression, forming a one-way mapping table from non-standard to standard expressions. This mapping table is used to clean the original dataset. Using the cleaned standardized knowledge points as the index primary key, the same platform ID is counted only once, regardless of how many courses it corresponds to, thus constructing a one-to-one correspondence index structure between standardized knowledge points and platform sets. A threshold for duplicate detection is set; when a knowledge point exists on two or more platforms, it is determined to be a duplicate knowledge point. All index entries are traversed, and the number of elements in the platform set corresponding to each standardized knowledge point is calculated one by one to obtain the counting result. The counting result of each knowledge point is compared with the preset threshold, and knowledge points whose counting results meet the threshold condition are selected to form a set of duplicate knowledge points.

[0027] Furthermore, for each repetitive knowledge point, teaching-related text data corresponding to that knowledge point is extracted from each platform; redundant information is removed from the teaching-related text data, and the text is segmented using the jieba word segmentation tool to form a structured text dataset corresponding to a single repetitive knowledge point on each platform; the structured text datasets from each platform are truncated or padded according to the model requirements and input into the BERT model; the bidirectional Transformer architecture of the BERT model is used to capture the contextual semantic relationships of core words in the text, and the text is transformed into a fixed-dimensional semantic feature vector. The K-Means unsupervised clustering algorithm is used to mine topics from the semantic feature vectors of individual repetitive knowledge points. Specifically, the optimal number of clusters is determined based on the elbow rule; the semantic feature vectors corresponding to each platform are input into the K-Means algorithm, which iteratively calculates and minimizes the sum of squares within each cluster, aggregating semantic feature vectors with high semantic similarity into the same topic cluster, with each topic cluster corresponding to a potential knowledge focus direction; after clustering, keywords are extracted from the text fragments within each topic cluster to generate topic tags. For each topic cluster to which the text data belongs on each platform, the TF-IDF algorithm is used to calculate the weight of each topic cluster. Specifically, taking the platform's text dataset as a unit, the word frequency (TF) of the text fragment corresponding to each topic cluster in the platform's dataset is counted. Combined with the inverse document frequency (IDF) of the keywords of the topic cluster in all platform datasets, the TF-IDF weight value of each topic cluster is calculated. The topic cluster with the highest weight value is selected, and its corresponding topic tag is determined as the platform's focus on the knowledge direction of this repetitive knowledge point.

[0028] Furthermore, the repetitive knowledge points, corresponding platforms, and key knowledge directions are linked together. The constituent dimensions of each key knowledge direction are broken down to establish a multi-dimensional quantitative evaluation system, clarifying the evaluation connotation and value range definition rules for each dimension. For each repetitive knowledge point and its corresponding platform, basic data for each evaluation dimension are extracted from teaching text datasets and course structure information. A normalization mathematical algorithm is used to map the basic data of different dimensions to a unified numerical range, eliminating evaluation bias caused by differences in the units of measurement between different dimensions, and obtaining standardized scores for each dimension. Based on the influence weight of each evaluation dimension on the degree of emphasis, a weight allocation algorithm is used to determine the weight coefficient of each dimension. A weighted aggregation algorithm is then used to calculate the weighted average of the standardized scores of each dimension with their corresponding weight coefficients, resulting in a comprehensive quantitative value of the degree of emphasis on the repetitive knowledge point for a single platform. The comprehensive quantitative values ​​of adjacent repetitive knowledge points within the same platform are logically compared, and the distinguishability of the quantitative results of the same repetitive knowledge point across different platforms is verified to eliminate outliers caused by data extraction bias or unreasonable weight allocation.

[0029] In one specific embodiment, based on the multi-platform knowledge point database (containing 1268 independent knowledge points and 2135 sets of relationships) formed in step S100, duplicate knowledge points are screened. By comparing the basic attributes of the knowledge points to determine their intrinsic equivalence, 12 non-standard expressions, such as "Python list slicing" and "Python list slicing syntax" and "list slicing operation methods," are reduced to the same standard expression, forming a one-way mapping table containing 86 sets of mapping relationships. After cleaning the original dataset using this table, 1052 standardized knowledge points are obtained. A "knowledge point-platform set" index is constructed using the standardized knowledge points as the primary key. The same platform ID is counted only once. The number of elements in the platform set is calculated by traversing all index items, and a threshold of "≥2" is set. Finally, 328 duplicate knowledge points are selected, accounting for 31.2% of the total number of standardized knowledge points. Typical examples include "Python list slicing," "for loop syntax," and "dictionary key-value pair operations."

[0030] The repetitive knowledge point "Python list slicing" was selected as a typical sample, and teaching-related text data corresponding to three platforms, A, B, and C, were extracted: 8200 words were extracted from platform A, 7500 words from platform B, and 6800 words from platform C. After removing redundant information, 6500 words, 5800 words, and 5200 words of effective text were obtained respectively. After jieba word segmentation and stop word filtering, a structured text dataset was formed. After processing into fixed-length text according to the requirements of the BERT model, it was input into the model and transformed into a 768-dimensional semantic feature vector.

[0031] The K-Means algorithm was used to mine topics from three sets of semantic feature vectors for this knowledge point. The elbow rule was used to determine the optimal number of clusters as 3. After iterative clustering, three topic clusters were formed, and keywords were extracted to generate topic tags: "Practical Steps Breakdown," "Theoretical Principle Explanation," and "Application of Real Exam Cases." The TF-IDF algorithm was used to calculate the weights of topic clusters on each platform: Platform A, "Practical Steps Breakdown," had a weight of 0.82, representing the core focus; Platform B, "Theoretical Principle Explanation," had a weight of 0.79, representing the core focus; and Platform C, "Application of Real Exam Cases," had a weight of 0.85, representing the core focus.

[0032] For "Python list slicing" and three platforms, a quantitative evaluation system was established, comprising three dimensions: "content proportion," "detailedness," and "case fit." Basic data was extracted: Platform A had "68% textual proportion of practical steps," "12 detailed description items," and "8 practical cases," with standardized scores of 8.5, 9.2, and 8.8 respectively after normalization. Weights were assigned based on the importance of each dimension (content proportion 0.4, detailedness 0.3, case fit 0.3), and a weighted average was calculated to obtain a comprehensive quantitative value of 8.8 for Platform A. Similarly, the comprehensive quantitative value for Platform B was calculated to be 7.6, and for Platform C, 9.1. Logical comparison and verification showed that the quantitative results for the three platforms exhibited good differentiation and no outliers.

[0033] Step S300: Based on the student's historical test data, analyze the student's performance on each knowledge point and deduce the student's emphasis on each knowledge point; through a quantitative evaluation model, convert the degree of the student's emphasis on each knowledge point into a quantitative value; compare the quantitative results of the student's emphasis on repetitive knowledge points with those of each platform one by one, and recommend the course with the closest degree of emphasis to the student. Specifically, the process involves collecting students' historical test data, cleaning the data, and removing invalid data; based on the historical test data, calculating the percentage of correct answers for each student across all questions related to that knowledge point, reflecting their overall mastery of the knowledge point; categorizing and statistically analyzing incorrect answers for that knowledge point according to pre-defined error types to identify the core causes of errors; calculating the average time spent answering questions for that knowledge point, and combining this with the time taken based on the difficulty level of the questions to determine the student's understanding and proficiency of the knowledge point; and comprehensively forming multi-dimensional statistical results of each student's performance across all knowledge points. The study systematically analyzes the types of knowledge points that emphasize specific knowledge directions, establishes a system of rules linking answer performance dimensions with the knowledge directions emphasized, and clarifies the emphasis requirements corresponding to different combinations of answer performance. It then matches the statistical results of each student's multi-dimensional answer performance for each knowledge point with the pre-defined system of rules. Finally, it employs a weighted judgment algorithm to rank the matched potential emphasis directions by weight, selecting the direction with the highest weight as the student's emphasis on that knowledge point.

[0034] Furthermore, based on the dimensions of individual students and individual repetitive knowledge points, the student's emphasis on that knowledge point is quantified and compared one by one with the quantified values ​​of the corresponding knowledge points on each platform. The platforms are then sorted by the degree of difference to determine the degree of matching between each platform and the student. For each repetitive knowledge point, the platform courses that are closest to the student's emphasis are selected. The best matching courses for all knowledge points are then compiled, and after removing duplicate courses, a personalized recommendation list is formed, thus completing the recommendation output.

[0035] In one specific embodiment, 10 Python programming beginners were selected as experimental subjects, and complete historical data from their last three Python basic tests were collected, including test papers, questions, correct and incorrect answers, and time spent answering questions. The data was cleaned, removing one missing test paper and two test papers with abnormal answering times (less than 50% of the standard time), ultimately retaining 27 valid test data. Focusing on the repetitive knowledge point "Python list slicing," the performance of the 10 students was analyzed: Student A answered 2 out of 5 questions related to this knowledge point correctly, a 40% accuracy rate; all incorrect answers were categorized as "omission of practical steps," accounting for 100%; the average answering time was 120 seconds, exceeding the benchmark time (80 seconds) by 50%. This resulted in a multi-dimensional statistical conclusion of "low accuracy rate, predominance of practical errors, and excessively long answering time."

[0036] Three key knowledge areas were identified: "Practical Step Breakdown," "Theoretical Principle Explanation," and "Application of Past Exam Questions." Correlation rules were established between answer performance and these areas (e.g., "low accuracy rate + ≥80% of practical questions answered incorrectly" corresponds to the "Practical Step Breakdown" area). Student A's answer performance statistics were matched against these correlation rules, resulting in two potential areas: "Practical Step Breakdown" and "Application of Past Exam Questions." After weighting using an algorithm, the "Practical Step Breakdown" area (0.91 weight) was higher than the "Application of Past Exam Questions" area (0.45 weight), thus identifying it as Student A's core focus for this knowledge point.

[0037] Using the quantitative evaluation model established in step S200, the emphasis of Student A's "practical step breakdown" is converted into a quantitative value of 8.7. Based on the dimension of "Student A - Python list slicing," the quantitative values ​​(8.8, 7.6, 9.1) for this knowledge point on platforms A, B, and C are compared one by one, and the degree of difference is calculated: 0.1 with platform A, 1.1 with platform B, and 0.4 with platform C. Based on the degree of difference, platform A is determined to have the highest matching closeness, and the "Python list slicing practical explanation" course on platform A is selected as the optimal matching course for this knowledge point.

[0038] Following the above process, all repetitive knowledge points for 10 students were compared and matched. The best-matching courses for each knowledge point were summarized, and two duplicate courses were removed. A personalized recommendation list was then generated for each student. Student A's recommendation list included 3 courses from platform A and 1 course from platform C. The matching accuracy of the list was manually verified to reach 92%, achieving accurate recommendation output.

[0039] Step S400: When multiple courses are consistent with the student's focus, construct a multi-dimensional quantitative scoring system, assign corresponding weights to each indicator and calculate the comprehensive score of each course; sort the courses from high to low according to the comprehensive score, and recommend the sorted courses to the corresponding students.

[0040] Specifically, indicators affecting course quality are selected and included in the multi-dimensional quantitative scoring system, including teacher qualifications, course approval ratings, and total course duration. Based on the degree of influence of each indicator on course quality, a weighting allocation logic is used to determine the weight coefficient of each type of indicator, ensuring that the total weight is 1. The original data of the corresponding indicators for each course are extracted and standardized. The standardized data is then weighted and aggregated with the corresponding indicator weights to obtain the comprehensive score for each course.

[0041] In one specific embodiment, when matching the knowledge point of "for loop syntax" for student A, two courses, "Python for loop practical tutorial" on platform A and "Python loop structure in detail" on platform C, were selected. The quantitative difference between the two courses and student A's focus (practical steps breakdown) on this knowledge point is 0.2, indicating that the focus is consistent. Further screening is required through comprehensive scoring.

[0042] A multi-dimensional quantitative scoring system was constructed, with teacher qualifications, course satisfaction rating, and total course duration as core indicators. Weights were assigned based on the degree of influence of each indicator on course quality, with teacher qualifications having a weight of 0.4, course satisfaction rating having a weight of 0.3, and total course duration having a weight of 0.3, for a total weight of 1.

[0043] Raw data for the metrics of two courses were extracted: Course A's instructor has 8 years of teaching experience and holds an advanced Python development certification; the course has a 96% positive review rate and a total duration of 45 minutes. Course C's instructor has 5 years of teaching experience and holds an intermediate Python development certification; the course has a 92% positive review rate and a total duration of 38 minutes. After standardizing the raw data, the standardized scores for each metric on platform A are 9.2, 9.6, and 8.5; for platform C, they are 7.5, 8.4, and 7.8.

[0044] The comprehensive score was calculated using weighted aggregation: Course A's comprehensive score = 9.2 × 0.4 + 9.6 × 0.3 + 8.5 × 0.3 = 9.09; Course C's comprehensive score = 7.5 × 0.4 + 8.4 × 0.3 + 7.8 × 0.3 = 7.86. Courses were sorted from highest to lowest based on their comprehensive scores, with Course A ranked first. These Courses were then included in Student A's personalized recommendation list, completing the precise ranking and recommendation process.

[0045] like Figure 2 As shown, this application provides a data analysis-based intelligent management system for learning resources, including: The knowledge graph construction module includes: a data acquisition unit that traverses all courses on various online education platforms and extracts the knowledge points covered by each course; a single-course knowledge graph construction unit that, for a single course, sorts out the progressive and horizontal relationships between knowledge points and constructs the knowledge graph of the course; and a multi-platform knowledge point database aggregation unit that aggregates the knowledge graphs of all courses on all platforms to form a multi-platform knowledge point database. The module for analyzing and quantifying repetitive knowledge points includes: a repetitive knowledge point screening unit that filters repetitive knowledge points existing on two or more platforms based on a multi-platform knowledge point database; a platform-emphasized knowledge direction analysis unit that analyzes the emphasis of different platforms on each repetitive knowledge point; and a quantification unit that establishes a quantitative evaluation model to convert the degree of emphasis of each platform on the repetitive knowledge point into specific numerical values. The course matching module includes: a student-focused knowledge direction deduction unit that analyzes students' historical test data to break down their performance on each knowledge point and deduce their focus on each knowledge point; a student-focused degree quantification unit that uses a quantitative evaluation model to convert the degree of students' focus on each knowledge point into a quantitative value; and a quantitative result comparison and matching unit that compares the quantitative results of students' focus on repetitive knowledge points with those of various platforms one by one and recommends courses with the closest focus to students. The course recommendation optimization module includes: a course comprehensive score calculation unit that constructs a multi-dimensional quantitative scoring system when multiple courses align with a student's focus, assigns corresponding weights to each indicator, and calculates the comprehensive score for each course; and a course ranking and recommendation unit that ranks courses from highest to lowest comprehensive score and recommends the ranked courses to the corresponding students.

[0046] 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 data analysis-based intelligent management method for learning resources, characterized in that, Includes the following steps: Iterate through all courses on various online education platforms and extract the knowledge points covered in each course; for a single course, sort out the progressive and horizontal relationships between knowledge points and construct the knowledge graph of the course; summarize the knowledge graphs of all courses on all platforms to form a multi-platform knowledge point database; Based on a multi-platform knowledge point database, we screened out repetitive knowledge points that exist on two or more platforms; for each repetitive knowledge point, we analyzed the emphasis of different platforms in explaining it; and we established a quantitative evaluation model to convert the degree of emphasis of each platform on the repetitive knowledge point into specific values. Based on students' historical test data, we analyze their performance on each knowledge point and deduce their emphasis on each knowledge point. Through a quantitative evaluation model, we convert the degree of students' emphasis on each knowledge point into a quantitative value. We then compare the quantitative results of students' emphasis on repetitive knowledge points with those of various platforms and recommend courses with the closest degree of emphasis to students. When multiple courses align with a student's focus, a multi-dimensional quantitative scoring system is constructed, assigning corresponding weights to each indicator and calculating a comprehensive score for each course. Courses are then ranked from highest to lowest based on their comprehensive scores, and the ranked courses are recommended to the corresponding students.

2. The intelligent management method for learning resources based on data analysis according to claim 1, characterized in that, The process involves iterating through all courses on various online education platforms and extracting the knowledge points covered in each course, including: Identify the online education platforms to be covered, define the types and status of courses to be processed on each platform, and determine the traversal boundaries; obtain the basic information of all courses on each platform in batches through platform open interface calls and standardized data crawling; for platforms without open interfaces, use automated scripts to traverse the course list page according to preset rules, locate and obtain the basic information of individual courses one by one; For each course, its core teaching resources are analyzed, redundant information is removed, and text and structured content related to the core teaching are extracted to form a set of teaching content materials for a single course. Based on the teaching content material set, knowledge points with independent teaching significance are identified and separated through semantic analysis, keyword clustering, and teaching logic decomposition. For content with hierarchical relationships, it is first separated into the finest-grained independent knowledge points, and then its level is recorded. The extracted independent knowledge points are initially screened for repetition, and the accuracy and completeness of the knowledge point descriptions are verified. Ambiguous, unclear, or content without actual teaching significance is removed to form a list of independent knowledge points for a single course.

3. The intelligent management method for learning resources based on data analysis according to claim 1, characterized in that, For a single course, the progressive and horizontal relationships between knowledge points are identified, and a knowledge graph for the course is constructed, including: Each independent knowledge point in a single course is labeled with basic attributes; based on the basic attributes of the knowledge points, the course teaching logic, cognitive learning patterns, and difficulty levels of the knowledge points are extracted, the progressive dependencies between knowledge points are sorted out, the order of learning knowledge points first and later is clarified, the support relationship between basic knowledge points and advanced knowledge points is determined, and a list of progressive relationships is formed; based on the basic attributes of the knowledge points, the knowledge point types, teaching application scenarios, and the logic of the after-class exercises are extracted, the horizontal connections between knowledge points at the same level or across levels are sorted out, different knowledge points for solving the same problem and collaborative knowledge points in the same application scenario are identified, the connection dimensions between each knowledge point are clarified, and a list of horizontal connection relationships is formed. Using independent knowledge points as nodes and the progressive relationships as edges, a structured knowledge graph for a single course is constructed, and the basic attributes and relationship descriptions of each node are marked in the graph.

4. The intelligent management method for learning resources based on data analysis according to claim 1, characterized in that, The multi-platform knowledge point database filters out repetitive knowledge points existing on two or more platforms, including: Based on a multi-platform knowledge point database, this method compares the basic attributes of knowledge points to determine the intrinsic equivalence between different expressions. All equivalent non-standard expressions are then reduced to the same standard expression, forming a one-way mapping table from non-standard to standard expressions. This mapping table is used to clean the original dataset. Using the cleaned standardized knowledge points as the primary index key, the same platform ID is counted only once, regardless of how many courses it corresponds to, thus constructing a one-to-one correspondence index structure between standardized knowledge points and platform sets. A threshold for duplicate detection is set; when a knowledge point exists on two or more platforms, it is considered a duplicate knowledge point. All index entries are traversed, and the number of elements in the platform set corresponding to each standardized knowledge point is calculated to obtain the counting result. The counting result for each knowledge point is compared with the preset threshold, and knowledge points whose counting results meet the threshold condition are selected to form a set of duplicate knowledge points.

5. The intelligent management method for learning resources based on data analysis according to claim 1, characterized in that, For each repetitive knowledge point, the analysis of the emphasis placed on different knowledge areas in the explanations on different platforms includes: For each repetitive knowledge point, teaching-related text data corresponding to that knowledge point is extracted from each platform; redundant information is removed from the teaching-related text data, and the text is segmented using the jieba word segmentation tool to form a structured text dataset corresponding to a single repetitive knowledge point on each platform; the structured text datasets from each platform are truncated or padded according to the model requirements and input into the BERT model; the bidirectional Transformer architecture of the BERT model is used to capture the contextual semantic relationships of core words in the text, and the text is transformed into a fixed-dimensional semantic feature vector; The K-Means unsupervised clustering algorithm is used to mine topics from the semantic feature vectors of individual repetitive knowledge points. Specifically, the optimal number of clusters is determined based on the elbow rule; the semantic feature vectors corresponding to each platform are input into the K-Means algorithm, which iteratively calculates and minimizes the sum of squares within each cluster, aggregating semantic feature vectors with high semantic similarity into the same topic cluster, with each topic cluster corresponding to a potential knowledge focus direction; after clustering, keywords are extracted from the text fragments within each topic cluster to generate topic tags. For each topic cluster to which the text data belongs on each platform, the TF-IDF algorithm is used to calculate the weight of each topic cluster. Specifically, taking the platform's text dataset as a unit, the word frequency (TF) of the text fragment corresponding to each topic cluster in the platform's dataset is counted. Combined with the inverse document frequency (IDF) of the keywords of the topic cluster in all platform datasets, the TF-IDF weight value of each topic cluster is calculated. The topic cluster with the highest weight value is selected, and its corresponding topic tag is determined as the platform's focus on the knowledge direction of this repetitive knowledge point.

6. The intelligent management method for learning resources based on data analysis according to claim 1, characterized in that, The establishment of a quantitative evaluation model converts the degree of emphasis placed on this repetitive knowledge point by each platform into specific numerical values, including: By associating repetitive knowledge points with corresponding platforms and key knowledge areas, and breaking down the constituent dimensions of each key knowledge area, a multi-dimensional quantitative evaluation system is established, clarifying the evaluation connotation and value range definition rules for each dimension. For each repetitive knowledge point and its corresponding platform, basic data for each evaluation dimension is extracted from teaching text datasets and course structure information. A normalization mathematical algorithm is used to map the basic data of different dimensions to a unified numerical range, eliminating evaluation bias caused by differences in the units of measurement between different dimensions, and obtaining standardized scores for each dimension. Based on the influence weight of each evaluation dimension on the degree of emphasis, a weight allocation algorithm is used to determine the weight coefficient of each dimension. A weighted aggregation algorithm is then used to calculate the weighted average of the standardized scores of each dimension with their corresponding weight coefficients, resulting in a comprehensive quantitative value of the degree of emphasis on the repetitive knowledge point for a single platform. The comprehensive quantitative values ​​of adjacent repetitive knowledge points within the same platform are logically compared, and the distinguishability of the quantitative results of the same repetitive knowledge point across different platforms is verified to eliminate outliers caused by data extraction bias or unreasonable weight allocation.

7. The intelligent management method for learning resources based on data analysis according to claim 1, characterized in that, Based on students' historical test data, the analysis breaks down students' performance on various knowledge points, deduce their emphasis on each knowledge point, including: Collect students' historical test data, clean the data, and remove invalid data; based on the historical test data, calculate the percentage of correct answers for each student across all questions corresponding to that knowledge point, reflecting the overall mastery of the knowledge point; categorize and statistically analyze the incorrect questions for that knowledge point according to preset error types, and identify the core causes of errors; calculate the average answering time for each question corresponding to that knowledge point, and combine this with the time taken based on the difficulty level of the questions to determine the student's understanding and proficiency of the knowledge point; comprehensively form multi-dimensional statistical results of each student's answering performance for each knowledge point; The study systematically analyzes the types of knowledge points that emphasize specific knowledge directions, establishes a system of rules linking answer performance dimensions with the knowledge directions emphasized, and clarifies the emphasis requirements corresponding to different combinations of answer performance. It then matches the statistical results of each student's multi-dimensional answer performance for each knowledge point with the pre-defined system of rules. Finally, it employs a weighted judgment algorithm to rank the matched potential emphasis directions by weight, selecting the direction with the highest weight as the student's emphasis on that knowledge point.

8. The intelligent management method for learning resources based on data analysis according to claim 1, characterized in that, The process involves comparing the quantitative results of students' repetitive knowledge point emphases with those of various platforms one by one, and recommending courses with the closest emphases to students, including: Based on the dimensions of individual students and individual repetitive knowledge points, the student's emphasis on that knowledge point is quantified and compared with the quantified values ​​of the corresponding knowledge points on each platform. The platforms are then sorted by the degree of difference to determine the degree of matching between the students and each platform. For each repetitive knowledge point, the platform courses that are closest to the student's emphasis are selected. The best matching courses for all knowledge points are compiled, and after removing duplicate courses, a personalized recommendation list is formed, thus completing the recommendation output.

9. The intelligent management method for learning resources based on data analysis according to claim 1, characterized in that, When multiple courses align with a student's area of ​​focus, a multi-dimensional quantitative scoring system is constructed, assigning corresponding weights to each indicator and calculating a comprehensive score for each course, including: Indicators affecting course quality were selected and incorporated into a multi-dimensional quantitative scoring system, including teacher qualifications, course ratings, and overall course duration. Based on the degree of influence of each indicator on course quality, a weighting allocation logic was used to determine the weight coefficient of each type of indicator, ensuring that the total weight is 1. The original data of the corresponding indicators for each course were extracted and standardized. The standardized data and the corresponding indicator weights were then weighted and aggregated to obtain the comprehensive score for each course.

10. A data analysis-based intelligent management system for learning resources, using the data analysis-based intelligent management method for learning resources according to any one of claims 1-9, characterized in that, include: The knowledge graph construction module includes: a data acquisition unit that traverses all courses on various online education platforms and extracts the knowledge points covered by each course; a single-course knowledge graph construction unit that, for a single course, sorts out the progressive and horizontal relationships between knowledge points and constructs the knowledge graph of the course; and a multi-platform knowledge point database aggregation unit that aggregates the knowledge graphs of all courses on all platforms to form a multi-platform knowledge point database. The module for analyzing and quantifying repetitive knowledge points includes: a repetitive knowledge point screening unit that filters repetitive knowledge points existing on two or more platforms based on a multi-platform knowledge point database; a platform-emphasized knowledge direction analysis unit that analyzes the emphasis of different platforms on each repetitive knowledge point; and a quantification unit that establishes a quantitative evaluation model to convert the degree of emphasis of each platform on the repetitive knowledge point into specific numerical values. The course matching module includes: a student-focused knowledge direction deduction unit that analyzes students' historical test data to break down their performance on each knowledge point and deduce their focus on each knowledge point; a student-focused degree quantification unit that uses a quantitative evaluation model to convert the degree of students' focus on each knowledge point into a quantitative value; and a quantitative result comparison and matching unit that compares the quantitative results of students' focus on repetitive knowledge points with those of various platforms one by one and recommends courses with the closest focus to students. The course recommendation optimization module includes: a course comprehensive score calculation unit that constructs a multi-dimensional quantitative scoring system when multiple courses align with a student's focus, assigns corresponding weights to each indicator, and calculates the comprehensive score for each course; and a course ranking and recommendation unit that ranks courses from highest to lowest comprehensive score and recommends the ranked courses to the corresponding students.

Images