Middle and primary school student knowledge mastering evaluation intelligent learning and practice task generation algorithm based on knowledge graph

Abstract

The invention discloses an intelligent learning and practice task generation algorithm for knowledge mastery evaluation of primary and secondary school students based on a knowledge graph, and relates to the technical field of educational informationization, and the algorithm comprises the following steps: comprehensively obtaining multi-dimensional original data in the whole process of executing personalized learning and practice task scheduling in the knowledge graph, changing the knowledge graph from a static structure to a dynamic observation operation state; the method comprises the following steps: preprocessing obtained multi-dimensional original data to obtain a data frame structure with a unified dimension; according to the method, knowledge graph operation data is dynamically perceived, the structural defect degree is quantified, an urgent area is positioned and connected based on the knowledge flow potential energy difference, a transition knowledge unit is generated through intelligent reasoning, an edge structure is expanded, a new path conforming to cognitive logic is reconstructed, the problems of graph chain breakage, islands and the like are solved, the task scheduling continuity and stability are improved, and the task scheduling efficiency is improved. And accurate pushing and cognitive adaptation of the personalized learning practice tasks are realized.

Description

Technical Field

[0001] The present invention relates to the field of educational informatization technology, and specifically to an algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on a knowledge graph. Background Art

[0002] "Intelligent learning and practice task generation for primary and secondary school students' knowledge mastery assessment based on knowledge graphs" refers to the use of knowledge graphs to systematically model the logical structure and hierarchical relationships between the knowledge points learned by primary and secondary school students in various subjects. This system combines students' learning data from daily homework, tests, and classroom activities, and uses multi-dimensional features (such as answer time and error types) to probabilistically assess the mastery of each knowledge point. Based on this assessment result, artificial intelligence algorithms such as reinforcement learning are used to automatically filter out highly relevant learning and practice content that matches their cognitive level from the knowledge graph, dynamically generating personalized learning task packages. This task package not only covers basic consolidation and ability expansion, but also features intelligent recommendations and adaptive adjustment features, achieving closed-loop optimization from knowledge diagnosis, personalized training, and learning feedback, thereby improving learning efficiency and teaching quality and meeting the differentiated development needs of students.

[0003] Existing technologies have the following shortcomings: In existing knowledge graph-based personalized learning and practice task generation systems, knowledge graphs are typically composed of multiple nodes representing specific knowledge points and the cognitive dependencies between them. However, during the actual construction and maintenance process, due to curriculum adjustments, graph version upgrades, or manual modeling omissions, structural defects may exist in the graph. For example, some nodes are in a "structural island" state, that is, they have no valid semantic association edges with any other knowledge points, or some key path edges that should have existed are mistakenly deleted, blocked, or omitted, resulting in "broken links" in the graph structure.

[0004] In this case, when the system invokes the reinforcement learning algorithm to generate a task path based on the student's knowledge mastery assessment results, the algorithm may be unable to retrieve a continuous path from the current knowledge state to the target knowledge point in the graph. To maintain the coherence of the generated path, the policy model may mistakenly forcibly splice two knowledge points with a large cognitive span and no actual cognitive transition into a continuous task node without a clear logical connection, resulting in structural jumps and cognitive gaps in the learning path.

[0005] This problem not only destroys the logical progression between knowledge tasks, but also seriously affects the adaptability and comprehensibility of learning tasks. It is very easy to cause students to have comprehension barriers, repeated errors and even frustration in the actual process of completing learning and practice tasks, thereby weakening the effectiveness and credibility of the intelligent teaching system.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide an intelligent learning and practice task generation algorithm for knowledge mastery assessment of primary and secondary school students based on knowledge graphs. By dynamically perceiving the knowledge graph operation data and quantifying the degree of structural defects, it locates the urgent connection areas based on the "knowledge flow potential difference", generates transitional knowledge units through intelligent reasoning and expands the edge structure, reconstructs new paths that conform to cognitive logic, solves problems such as graph breakage and islands, improves the consistency and stability of task scheduling, and realizes the accurate push and cognitive adaptation of personalized learning and practice tasks, so as to solve the problems in the above-mentioned background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graphs, comprising the following steps: During the entire process of executing personalized learning and training task scheduling on the knowledge graph, multi-dimensional raw data is fully acquired, transforming the knowledge graph from a static structure to a dynamic observation operation state; Preprocess the acquired multi-dimensional raw data to obtain a data frame structure with unified dimensions; Using feature engineering methods, we extract key features that characterize the structural defects of the knowledge graph from the preprocessed data, and conduct a comprehensive analysis of the extracted key features to quantify the severity of the structural defects of the knowledge graph. The quantified key features are fed into a pre-trained machine learning model. The model identifies the patterns and distribution of structural anomalies and assesses in real time whether there are structural defects in the knowledge graph operation. When structural defects in the graph are identified, a "knowledge flow potential difference" indicator is constructed to measure the urgency of connection based on the semantic relevance of knowledge nodes, cognitive dependency order and student behavioral characteristics. Guided by the "knowledge flow potential difference" indicator, transitional knowledge units are inferred and generated in the corresponding area, the edge structure is dynamically expanded, and new connection paths that conform to cognitive logic are constructed to adjust the local connectivity state and structural layout of the graph.

[0009] Preferably, in the whole process of executing personalized learning and training task scheduling on the knowledge graph, in order to realize the transformation from static structure to dynamic observation operation state, the specific steps of data acquisition are as follows: During the student-task interaction stage, the learning platform collects students’ response behavior data in real time during the question-answering process; In the task path scheduling phase, the graph scheduling engine collects node call records, path access status, and local connectivity indicator structure response data; During the model execution phase, the model execution log is generated through the scheduling module and the path planning algorithm interface, forming a system-level execution tracking chain; During the data collection stage, multi-dimensional data is aggregated in real time to the knowledge graph operation monitoring center through a unified scheduling and monitoring module.

[0010] Preferably, a feature engineering method is used to extract key features that characterize the structural defects of the knowledge graph from the preprocessed data. The extracted features include the proportion of cognitive spans exceeding a reasonable range in the task sequence and the path reconstruction structure change rate of the knowledge subgraph within the continuous task scheduling cycle. The proportion of cognitive spans exceeding a reasonable range in the task sequence and the path reconstruction structure change rate of the knowledge subgraph within the continuous task scheduling cycle are comprehensively analyzed under the detection window to generate cognitive logic fracture indicators and local structure drift indicators respectively. The severity of the structural defects of the knowledge graph is quantified by the cognitive logic fracture indicators and the local structure drift indicators.

[0011] Preferably, the specific steps of comprehensively analyzing the proportion of cognitive spans exceeding a reasonable range in the task sequence under the detection window to generate a cognitive logic fracture index are as follows: In the scheduling process of personalized chemistry learning tasks, the threshold of cognitive span is set to , which is used to limit the cognitive acceptance interval between adjacent task nodes. For any set of consecutive task nodes, the cognitive span difference in the knowledge graph is obtained, which is marked as ,when When , it is regarded as a cognitive rupture event. To characterize the intensity of each rupture event, the cognitive rupture intensity factor is calculated. The calculation expression is as follows: , Where, It is The cognitive span difference between the group task node pairs, It is The spectral connection reachability marker factor of the group task node pair, It is The semantic embedding distance factor of the group task node pair, It is cognitive fracture intensity factor of group task node pairs; Under the detection window, suppose there are Group task node pairs, among which the ones that meet the break conditions are The cognitive jump events in all task paths were analyzed cumulatively to generate the cognitive logic fracture index. The calculation formula is as follows: , Where, is the total number of task fragments that have broken in the detection window, is the total number of task fragments, that is, the total number of task node pairs, It is Path structure buffer parameters for non-broken segments, It is an indicator of cognitive logic rupture.

[0012] Preferably, the specific steps of comprehensively analyzing the path reconstruction structure change rate of the knowledge subgraph within the continuous task scheduling cycle under the detection window to generate the local structure drift index are as follows: In the detection window, select the target knowledge subgraph and record the task path sets generated in several consecutive task scheduling cycles. ,in Indicates that the target knowledge subgraph is in The set of valid task paths actually constructed within a scheduling cycle, is the number of scheduling cycles. To measure the degree of dynamic change of the graph path structure, a path structure change factor is introduced to measure the reconstruction amplitude of the path set between two adjacent scheduling cycles. The calculation expression of the path structure change factor is as follows: , Where, is the path structure change factor, which indicates the difference between the two consecutive scheduling cycles. The scheduling cycle and The degree to which the task path structure of the target knowledge subgraph changes between scheduling cycles, The target knowledge subgraph is in The set of valid task paths actually constructed within a scheduling cycle, is the cardinality of the intersection of the path sets, is the cardinality of the union of the path sets; Based on the obtained structural change factors of all paths , the structural drift index is introduced to comprehensively characterize the volatility and intensity of structural changes. The local structural drift index is generated by weighted fusion of the difference amplitude and change intensity of the change factors in consecutive periods. The calculation expression is as follows: , Where, is the natural base, is the trend amplification factor, It is The path structure change factor of the target knowledge subgraph between a scheduling cycle and the next cycle, is an indicator of local structural drift.

[0013] Preferably, the quantified cognitive logic fracture index and local structural drift index are input into a pre-trained machine learning model. The model generates a structural defect risk coefficient by identifying the pattern and distribution of structural abnormality features, and evaluates in real time whether there are structural defects in the operation of the knowledge graph based on the structural defect risk coefficient.

[0014] Preferably, the structural defect risk coefficient generated when the pattern and distribution of structural abnormality characteristics are identified by the pre-trained machine learning model is compared with the pre-set structural defect risk coefficient reference threshold to evaluate in real time whether there are structural defects in the operation of the knowledge graph. The evaluation process is as follows: If the structural defect risk coefficient is greater than the pre-set structural defect risk coefficient reference threshold, it is judged that there is a structural defect in the operation of the knowledge graph; if the structural defect risk coefficient is less than or equal to the pre-set structural defect risk coefficient reference threshold, it is judged that there is no structural defect in the operation of the knowledge graph.

[0015] Preferably, after identifying structural defects in the graph, a "knowledge flow potential difference" indicator is constructed to measure the urgency of connection based on the semantic relevance of knowledge nodes, cognitive dependency order, and student behavior characteristics. Guided by the "knowledge flow potential difference" indicator, transitional knowledge units are inferred and generated in the corresponding area, and the edge structure is dynamically expanded to construct a new connection path that conforms to cognitive logic. The specific steps are as follows: When structural defects are identified in the knowledge graph, a comprehensive evaluation is performed on the node characteristics of the defect-affected area to construct a regional weight factor that reflects the importance of the local structure and the urgency of repair. The calculation formula for the structural defect regional weight factor is as follows: , Where, is the average semantic relatedness, is the average cognitive dependency order weight, is the average student behavior trigger frequency, The average semantic relatedness , average cognitive dependency order weight and average student behavioral trigger frequency The weight coefficient of is the structural defect area weight factor; Based on the obtained structural defect area weight factor , combined with the severity of structural defects, the "knowledge flow potential difference" indicator is calculated by the ratio of the current structural defect risk coefficient to the set reference threshold. It is used to characterize the structural tension and connectivity requirements caused by the lack of connections in the current graph. The calculation expression of the "knowledge flow potential difference" indicator is as follows: , Where, is the structural defect risk factor, is the reference threshold of the structural defect risk coefficient, It is the potential energy difference of knowledge flow; Knowledge flow potential difference To drive, inference is generated in the structural missing area, and the knowledge flow potential difference is generated. The numerical strength of dynamically expands the edge structure, builds new connection paths, reconstructs the cognitive chain, and realizes the adaptive repair of the local structure of the atlas. The formula is as follows: , Where, It is a set of transitional knowledge units. is the transitional knowledge unit reasoning function, is the structural expansion magnification factor, is the edge structure extension.

[0016] In the above technical solution, the technical effects and advantages provided by the present invention are: Through continuous perception of multidimensional data and intelligent identification of structural defects during the operation of the knowledge graph, this invention can dynamically quantify the connectivity health of the graph. Based on the "knowledge flow potential difference" indicator, it locates areas of critical connectivity. It further infers and generates transitional knowledge units and implements edge structure expansion, thereby reconstructing a path structure that conforms to cognitive logic while maintaining semantic consistency. This mechanism effectively avoids task scheduling errors caused by graph disconnection, islands, or missing edges, improves the continuity and rationality of personalized task recommendations, and enhances the system's path generation stability and cognitive adaptability in real-world learning scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0018] Figure 1 This is a flow chart of the method for generating an algorithm for intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graphs in the present invention. DETAILED DESCRIPTION

[0019] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0020] The present invention provides Figure 1 The algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graphs includes the following steps: During the entire process of executing personalized learning and training task scheduling on the knowledge graph, multi-dimensional raw data is fully acquired, transforming the knowledge graph from a static structure to a dynamic observation operation state; In the knowledge graph-driven personalized learning and practice task scheduling process, the system not only relies on the pre-set graph structure to generate task paths but also needs to dynamically perceive the graph's response status during actual operation. Therefore, comprehensive data collection is required, including three key categories: ① Student task response behavior data, such as answering time, accuracy rate, and question skipping frequency, which reflect students' actual learning performance at each knowledge node; ② Graph structure response data, such as graph local connectivity, node access frequency, and path access success rate, which measure the accessibility and stability of the graph structure during task execution; and ③ Model execution log data, such as node jump sequences and path failure records in task planning, which are used to retrace algorithm decision paths and identify structural anomalies. This data can be collected in real time through front-end interaction log collection (such as the answering platform), back-end scheduling system monitoring modules, and the graph engine's path tracing mechanism. Aggregating this multi-dimensional raw data enables dynamic observation of the knowledge graph's operational status, transforming the graph from a static storage structure into a "measurable, controllable, and diagnosable" intelligent execution entity, providing a data foundation for subsequent structural defect identification and intelligent repair.

[0021] This step transforms the knowledge graph from a static structure to a dynamically observed operating state, ensuring that abnormal pattern data can be captured in real time, accurately, and comprehensively.

[0022] In the entire process of executing personalized learning and practice task scheduling on the knowledge graph, in order to achieve the transformation from static structure to dynamic observation operation state, the specific steps of data acquisition are as follows: First, in the student-task interaction stage, the learning platform collects students' response behavior data in the process of answering questions in real time, such as answering time, error rate, skipping frequency, etc., to record the actual interactive performance of knowledge points; second, in the task path scheduling stage, the graph scheduling engine collects node call records, path access status, and local connectivity indicator structure response data to reflect the accessibility and stability of the graph in the path generation process; third, in the model execution stage, the scheduling module and the path planning algorithm interface generate model execution logs, including node access sequences, path construction links and abnormal jump information during task recommendation, forming a system-level execution tracking chain; finally, in the data collection stage, the above multi-dimensional data are aggregated in real time to the knowledge graph operation monitoring center through a unified scheduling monitoring module, thereby building an observable and traceable knowledge graph dynamic behavior data system to support subsequent structural defect identification.

[0023] Preprocess the acquired multi-dimensional raw data to obtain a data frame structure with unified dimensions; Data cleaning and standardization are performed on the acquired raw behavior, structure, and log data, which mainly includes the following processes: First, duplicate record cleaning and missing value processing are performed. By identifying duplicate data entries, invalid samples, and missing fields, data integrity is improved by using strategies such as deletion, mean interpolation, or sliding window filling. Second, abnormal noise values ​​are eliminated, and rule discrimination (such as extreme value thresholds, logical conflicts) or statistical detection (such as box plots, Z-score analysis) is used to identify and process atypical behavior records to ensure the rationality of feature distribution. Next, timestamp alignment is performed to uniformly convert time records from multi-source data into a standard format and align them according to the task execution sequence to ensure temporal consistency. Subsequently, feature space alignment is completed, and data fields from different sources are mapped, naming standards are unified, and dimensions are converted so that they can be used in the same feature system. Finally, data normalization is applied to perform interval scaling on numerical features (such as Min-Max normalization or Z-score normalization) to improve the model's adaptability to feature differences. The purpose of this process is to convert raw data with inconsistent structure and unstable quality into standardized input that can be used for subsequent modeling, analysis and structural defect identification, thereby improving the data consistency, expression accuracy and processing efficiency of the overall system.

[0024] After preprocessing is completed, a data frame structure with unified dimensions is obtained, which enables subsequent feature engineering technology to efficiently perform data mining and feature extraction, thereby improving the accuracy and reliability of subsequent analysis and modeling.

[0025] Using feature engineering methods, we extract key features that characterize structural defects in the knowledge graph (such as structural islands, broken links, and missing edges) from the preprocessed data. We then conduct a comprehensive analysis of the extracted key features to quantify the severity of the structural defects in the knowledge graph. A feature engineering method is used to extract key features that characterize the structural defects of the knowledge graph (such as structural islands, broken links, and edge omissions) from the preprocessed data. The extracted features include the proportion of cognitive spans exceeding a reasonable range in the task sequence and the path reconstruction structure change rate of the knowledge subgraph within the continuous task scheduling cycle. The proportion of cognitive spans exceeding a reasonable range in the task sequence and the path reconstruction structure change rate of the knowledge subgraph within the continuous task scheduling cycle are comprehensively analyzed under the detection window to generate cognitive logic fracture indicators and local structure drift indicators, respectively. The cognitive logic fracture indicators and local structure drift indicators are used to quantify the severity of the structural defects of the knowledge graph.

[0026] The increasing proportion of task sequences with cognitive spans exceeding reasonable limits indicates that the system frequently skips necessary cognitive transition nodes when generating personalized learning paths, resulting in nonlinear cognitive jumps. This phenomenon directly indicates structural flaws in the knowledge graph. Under normal circumstances, a knowledge graph should ensure that task paths exhibit a cognitive progression from shallow to deep, from basic to complex—a phenomenon known as "cognitive continuity." When broken links, missing edges, or isolated key knowledge nodes appear in the graph, the system is unable to generate a reasonable path based on the original topology. Consequently, the scheduling mechanism is forced to forcibly splice task nodes with large logical spans, disrupting the progressive nature of the cognitive path. This task skipping is not due to individual student differences but rather the result of the graph structure's inability to provide a complete learning chain. Abnormally widened cognitive span is a passive manifestation of the system's attempt to "complete tasks" within structural constraints. Therefore, this feature not only indicates an abnormal student learning path but also directly reflects structural flaws in the graph at the behavioral level, demonstrating high sensitivity and diagnostic value.

[0027] The specific steps for comprehensively analyzing the proportion of cognitive spans exceeding the reasonable range in the task sequence under the detection window to generate cognitive logic fracture indicators are as follows: In the scheduling process of personalized chemistry learning tasks, the threshold of cognitive span is set to , which is used to limit the cognitive acceptance interval between adjacent task nodes. For any set of consecutive task nodes, the cognitive span difference in the knowledge graph is obtained, which is marked as , the difference can be calculated based on the number of path hops, cognitive level difference or structural level distance. When , it is regarded as a cognitive rupture event. To characterize the intensity of each rupture event, the cognitive rupture intensity factor is calculated. The calculation expression is as follows: , Where, It is The cognitive span difference of a pair of task nodes indicates the cognitive distance between two consecutive task nodes in the knowledge graph, reflecting the span between knowledge points in the current task path. It can be defined based on the structural distance in the knowledge graph (such as the shortest path length of the node), the cognitive level difference (such as the difference in the number of the knowledge stage to which it belongs), or the logical jump level in the knowledge point dependency graph. It is used to quantify the span of cognitive content between tasks. The larger the span, the lack of necessary cognitive connection between the two knowledge points. It is The spectral connection accessibility flag factor of the group task node pair is used to indicate whether the task node pair has a direct or indirect connection path in the knowledge graph. When there is no path between the node pairs (that is, there is a broken link, island or missing edge), ; When there is a valid connection between the node pairs (e.g., reachable through a multi-hop path or logical edge), , It is The semantic embedding distance factor of the task node pair is used to measure the semantic similarity difference between the task node pairs, that is, the semantic "distance" of the two knowledge points in the concept space. It can be obtained by calculating the Euclidean distance, cosine distance or Manhattan distance through the node embedding vector in the graph (such as the semantic embedding vector generated by models such as Word2Vec, Node2Vec, and GraphSAGE). The larger the semantic distance, the greater the conceptual difference of the knowledge points involved in the task jump, the higher the difficulty of knowledge transfer, and the increased cognitive burden on students. The smaller the span, the closer the semantics of the two knowledge points are. Even if the span is slightly larger, it is easier to understand. It is The cognitive fracture intensity factor of the group task node pair is used to characterize the severity of the structural fracture of the current task jump segment. It is a composite measurement result of comprehensive cognitive span, structural connectivity and semantic difference. If the cognitive span of a segment is large ( Large), unreachable in graph structure ( ) and the semantic difference is significant ( large), then A large value indicates that the path segment is a high-intensity fracture point; The fracture intensity factor comprehensively considers the amplitude of the span jump, the connectivity state of the graph structure, and the distance of the semantic relationship, and can effectively quantify the severity of each potential structural jump in the task path.

[0028] Under the detection window, suppose there are Group task node pairs, among which the ones that meet the break conditions are The cognitive jump events in all task paths were analyzed cumulatively to generate the cognitive logic fracture index. The calculation formula is as follows: , Where, is the total number of task fragments that have broken in the detection window, is the total number of task fragments, that is, the total number of task node pairs, which is the basic counting range in the entire analysis window. It is The path structure buffer parameter of the non-broken fragment measures whether the system has structural "redundancy" or "fault tolerance" characteristics in the non-broken path. The larger the value, the more robust the path structure and the more natural the cognitive transition. It is a cognitive logic fracture indicator, which represents a quantitative indicator of the structural defects of the graph formed after the intensity aggregation of jump segments whose cognitive span exceeds a reasonable range in all task paths within the detection window. This indicator comprehensively reflects the severity of the "cognitive jump" phenomenon caused by broken links, missing edges or missing transition nodes in the knowledge graph in the current scheduling cycle.

[0029] The above steps enhance the recognition ability of high-strength fracture behavior by weighted accumulation of the square of fracture strength. At the same time, a structural buffer term is introduced to suppress the influence of non-structural disturbances in the normal path, thereby improving the discriminant validity of the structural defect quantification results.

[0030] The cognitive logic fracture index, generated by comprehensively analyzing the proportion of cognitive spans in a task sequence that exceed a reasonable range within a detection window, indicates structural flaws in the knowledge graph. A smaller index value indicates a more complete graph structure. This index measures the proportion of cognitive spans in a task sequence that exceed a reasonable range within a detection window and assesses the coherence and rationality of path generation by quantifying the frequency of "cognitive jumps" within the path. If the graph structure is complete, the system can generate a progressive learning path based on the pre-dependencies and cognitive order between knowledge points, and the fracture index remains low. However, when the graph has structural flaws such as broken links, missing edges, or missing key transition nodes, the system cannot construct a smooth path and can only jump to knowledge points with larger spans, resulting in frequent cognitive logic fractures and an increase in this index.

[0031] If the rate of change in the path reconstruction structure of a knowledge subgraph within consecutive task scheduling cycles exhibits a periodic, non-stationary trend, it typically indicates structural flaws in the knowledge graph. This phenomenon reflects the system's inability to reliably reuse existing path structures during repeated scheduling, and instead frequently reconstructs task paths locally within the graph. This is typically due to broken links, missing edges, or structural islands in that region of the graph, forcing the scheduling strategy to adopt different alternative paths for compensatory connections in each cycle. Due to the lack of a fixed, optimal path structure, the scheduling path exhibits a periodic, fluctuating pattern of switching between several alternative paths, accompanied by high fluctuations in the structural drift rate and difficulty in convergence. This non-stationary structural drift trend indicates incomplete local connection logic and weak path accessibility within the knowledge graph. In severe cases, this can lead to cognitive jumps and task reconstruction failures. Therefore, the periodic non-stability of this indicator can serve as an important signal for identifying potential structural flaws in the graph and has strong diagnostic value.

[0032] The specific steps for comprehensively analyzing the path reconstruction structure change rate of the knowledge subgraph within the continuous task scheduling cycle under the detection window to generate the local structure drift index are as follows: In the detection window, select the target knowledge subgraph and record the task path sets generated in several consecutive task scheduling cycles. ,in Indicates that the target knowledge subgraph is in The set of valid task paths actually constructed within a scheduling cycle, is the number of scheduling cycles. To measure the degree of dynamic change of the graph path structure, a path structure change factor is introduced to measure the reconstruction amplitude of the path set between two adjacent scheduling cycles. The calculation expression of the path structure change factor is as follows: , Where, is the path structure change factor, which indicates the difference between the two consecutive scheduling cycles. The scheduling cycle and The degree to which the task path structure of the target knowledge subgraph changes between scheduling cycles, with a value range of 0 to 1. The target knowledge subgraph is in The set of valid task paths actually constructed within a scheduling cycle, is the cardinality of the intersection of the path sets, indicating that With the The number of repeated paths in the cycle, that is, the number of task paths shared by the two cycles, is used to count the number of identical paths in the two path sets. A larger value indicates a higher degree of path reuse and smaller structural changes; a smaller value indicates a low path reuse rate, indicating that the structure may be unstable or have lost edges. It is the cardinality of the union of the path sets, representing the total number of all different task paths in two cycles. A large union cardinality indicates a large number of path samples. A low intersection ratio indicates a greater intensity of change in the path structure. Adding 1 is used for bias adjustment to prevent division by zero errors when the path set is extremely small. Based on the obtained structural change factors of all paths , the structural drift index is introduced to comprehensively characterize the volatility and intensity of structural changes. The local structural drift index is generated by weighted fusion of the difference amplitude and change intensity of the change factors in consecutive periods. The calculation expression is as follows: , Where, is the natural base, Is the trend amplification coefficient, which is used to control the response intensity of the exponential weighting factor. Its value range is Adjust between It is The scheduling cycle and the cycle after it (i.e. cycles), the path structure change factor of the target knowledge subgraph, It is a local structure drift indicator used to quantify the instability of the structure reconstruction of the target knowledge subgraph under the detection window. The value range is 0-1. The larger the value, the more severe the structural drift and the more unstable the local structure.

[0033] The larger the local structural drift index, generated by comprehensively analyzing the rate of change in the knowledge subgraph's path reconstruction structure within successive task scheduling cycles within a detection window, the greater the magnitude of the topological shift in the paths in that region over time, indicating poor stability and the inability of the scheduling strategy to consistently reuse fixed paths. This high-frequency, non-convergent structural reconstruction often stems from structural defects within the graph, such as missing edges, isolated nodes, or broken links. This causes the path generation mechanism to repeatedly attempt to replace connections to maintain scheduling logic. Therefore, an increase in the local structural drift index can be considered a significant signal of potential structural anomalies in the graph. Conversely, a low value for this index indicates that the graph's local topology remains stable over multiple cycles, with strong path reusability. This indicates that the graph's connectivity is complete, scheduling is smooth, and generally free of structural defects.

[0034] The quantified key features are fed into a pre-trained machine learning model (e.g., random forest, XGBoost, graph convolutional neural network). The model identifies the patterns and distribution of structural anomalies and assesses in real time whether there are structural defects in the knowledge graph. The quantified cognitive logic fracture indicators and local structural drift indicators are input into a pre-trained machine learning model (such as random forest, XGBoost, and graph convolutional neural network). The model generates a structural defect risk coefficient by identifying the pattern and distribution of structural abnormality features, and based on the structural defect risk coefficient, it evaluates in real time whether there are structural defects in the operation of the knowledge graph.

[0035] A "pre-trained machine learning model" refers to an intelligent classification or regression model that has been trained and fine-tuned using historical datasets before the system is officially deployed for structural defect detection in knowledge graphs. This model is capable of identifying characteristic patterns of structural anomalies during graph operation and accurately predicting new data. In this application scenario, such models typically include Random Forest, eXtreme Gradient Boosting (XGBoost), and Graph Convolutional Neural Networks (GCNs). These models learn from labeled input samples (i.e., data containing both normal and abnormal structural labels) during the training phase, extracting nonlinear mappings between multidimensional features and targets. During training, the model continuously optimizes parameters (such as tree depth, weight update strategy, and node embedding method) to accurately judge complex input feature combinations (such as the coordinated changes in cognitive logic break indicators and local structural drift indicators). Ultimately, a prediction mechanism is formed that can automatically identify structural anomaly patterns and output a structural defect risk factor.

[0036] The greatest advantage of using pre-trained models lies in their generalization and real-time responsiveness. In practical applications, graph structural defects manifest in diverse and frequently changing forms. Relying solely on rule matching or manual thresholding makes it difficult to cover all anomalies. Pre-trained machine learning models, however, can perform comprehensive analysis and feature coupling inference on multi-dimensional input data without human intervention, identifying abnormal distribution patterns that are imperceptible to the naked eye but are representative in terms of statistics and spatial structure. For example, random forests enhance model robustness by running multiple decision trees in parallel, while XGBoost uses a gradient boosting strategy to finely characterize the influence of high-order feature combinations. Graph convolutional neural networks are particularly well-suited to the graph structure of knowledge graphs, learning structural stability or anomaly propagation patterns from the topology of node connections. Leveraging these pre-trained models, the system uses real-time cognitive fracture and structural drift indicators as input during each graph run. Combined with historical structural anomaly evolution patterns, it outputs a structural defect risk factor within milliseconds. This factor serves as the trigger for subsequent structural adjustment or repair strategies, enabling intelligent and dynamic health monitoring for large-scale knowledge graph systems.

[0037] The structural defect risk coefficient generated by the pre-trained machine learning model when identifying the pattern and distribution of structural abnormality features is compared with the pre-set structural defect risk coefficient reference threshold to evaluate in real time whether there are structural defects in the knowledge graph operation. The evaluation process is as follows: If the structural defect risk coefficient is greater than the pre-set structural defect risk coefficient reference threshold, it is judged that there is a structural defect in the operation of the knowledge graph; if the structural defect risk coefficient is less than or equal to the pre-set structural defect risk coefficient reference threshold, it is judged that there is no structural defect in the operation of the knowledge graph.

[0038] When structural defects are identified in the graph, a "knowledge flow potential difference" indicator is constructed to measure the urgency of connection based on the semantic relevance of knowledge nodes, cognitive dependency order, and student behavioral characteristics. Guided by this indicator, transitional knowledge units are inferred and generated in the corresponding area. The edge structure is dynamically expanded to construct new connection paths that conform to cognitive logic, thereby adjusting the local connectivity and structural layout of the graph. This step aims to quantify the urgency of the need for inter-node connectivity by constructing a "knowledge flow potential difference" metric when the system identifies structural defects in the knowledge graph (such as isolated islands, broken paths, and missing critical edges). This metric then serves as a foundation for intelligent graph repair and structural optimization. Specifically, the "knowledge flow potential difference" metric is an innovative measure that simulates the intensity of cognitive path demand. It comprehensively considers the semantic relevance between knowledge nodes (such as content similarity or conceptual nesting between knowledge points), the order of cognitive dependencies (such as whether one knowledge point is a prerequisite for another), and students' actual task performance (such as frequent question skipping, path interruptions, and task failure rates). Higher values ​​indicate a greater cognitive gap between two nodes caused by structural disruptions, and a greater urgency for path repair.

[0039] Based on the high-risk areas marked by the "knowledge flow potential difference", the system can generate transitional knowledge units through the knowledge reasoning mechanism, that is, fill in the missing concept nodes or transitional task points in the original path. These new nodes can be generated through large models, knowledge embedding reasoning, or graph neural network deduction, and have semantic rationality and cognitive accessibility. Subsequently, the system dynamically adds connecting edges to the graph, establishes reasonable connections between the new nodes and the relevant nodes in the original graph, and constructs a new path structure, thereby restoring the local connectivity of the knowledge graph at the structural level and ensuring the coherence and transition of the student's task path at the cognitive level. This step not only solves the path interruption problem caused by structural defects, but also establishes a data-driven, semantically aware, and dynamically evolving graph self-healing mechanism, which provides structural guarantees for the stability of task recommendations and cognitive continuity in intelligent teaching systems.

[0040] After identifying structural defects in the graph, we construct a "knowledge flow potential difference" indicator to measure the urgency of connection based on the semantic relevance of knowledge nodes, cognitive dependency order, and student behavioral characteristics. Guided by this indicator, we infer and generate transitional knowledge units in the corresponding area, dynamically expand the edge structure, and construct a new connection path that conforms to cognitive logic. The specific steps are as follows: After identifying structural defects in the knowledge graph, a comprehensive evaluation is conducted on the node features of the defect-affected area to construct a regional weight factor that reflects the importance of the local structure and the urgency of repair. This factor is composed of three key dimensions: the semantic relevance between knowledge points (indicating the similarity of conceptual content), the cognitive dependency order (indicating the order of knowledge points in the learning logic), and the behavioral characteristics of students during task scheduling (such as skipping rate and failure rate). The formula for calculating the structural defect regional weight factor is as follows: , Where, It is the average semantic relevance, which indicates the average similarity between knowledge nodes in the defect area in the semantic space. The semantic distance of node pairs is calculated through pre-trained language models or graph embedding models (such as Word2Vec, BERT, Node2Vec, etc.). The higher the similarity, the stronger the concept association. Common calculation methods include cosine similarity, inverse normalization of Euclidean distance, etc. is the average cognitive dependency order weight, which represents the average dependency strength between nodes in the defective area in the learning cognitive order. The sequence dependency matrix is ​​obtained through curriculum standards, teaching syllabus or expert annotations, and the strength of the cognitive chain violated by the interrupted path in the area is calculated. The higher the value, the stricter the sequence requirements of the knowledge chain in the area, and the greater the impact of the interruption on the cognitive path. The average student behavior trigger frequency indicates the frequency of abnormal behaviors exhibited by students in task scheduling within the structural defect area, including but not limited to the skipping rate, question failure rate, path interruption times, repeated task rollback, etc. The data sources are student interaction logs, task completion logs, and path execution records. The larger the value, the more direct the structural defects in the area have affected the learning process, and the higher the intervention priority. The average semantic relatedness , average cognitive dependency order weight and average student behavioral trigger frequency The weight coefficient satisfies , is the weight factor of the structural defect area, which represents the comprehensive weight score of the "structural importance" and "repair urgency" of a structural defect area in the atlas. The larger the value, the more critical the area is in the knowledge structure and the more sensitive the cognitive chain is. At the same time, it also shows more obvious abnormal behavior in the student's task execution and should be repaired first. The value range is 0-1; Based on the obtained structural defect area weight factor , combined with the severity of structural defects, the "knowledge flow potential difference" indicator is calculated by the ratio of the current structural defect risk coefficient to the set reference threshold. It is used to characterize the structural tension and connectivity requirements caused by the lack of connections in the current graph. The calculation expression of the "knowledge flow potential difference" indicator is as follows: , Where, is the structural defect risk coefficient, ranging from 0 to 1. is the reference threshold of the structural defect risk coefficient, It is the potential energy difference of knowledge flow, which indicates the degree of urgency of the connection between two or more knowledge nodes in the graph, that is, the "structural driving force" of the system to restore cognitive connectivity in a certain structural defect area. It is a positive real number. The larger the value, the stronger the urgency of repairing the connectivity structure. It is a direct driving parameter of the structural repair and path reconstruction strategy, used to determine whether to perform structural repair, the intensity of repair, and the scope of repair. This potential energy difference index comprehensively considers the degree of structural abnormality and the sensitivity of the local area of ​​the map at the semantic and cognitive levels, and is a direct basis for determining whether structural repair is needed and the intensity of the repair.

[0041] Knowledge flow potential difference To drive, inference is generated in the structural missing area, and the knowledge flow potential difference is generated. The numerical strength of dynamically expands the edge structure, builds new connection paths, reconstructs the cognitive chain, and realizes the adaptive repair of the local structure of the atlas. The formula is as follows: , Where, It is a set of transitional knowledge units, which means that after the system detects structural defects and calculates the potential difference of knowledge flow, it generates a set of "intermediary knowledge points" or "cognitive transition units" through the intelligent reasoning mechanism to fill the missing path. It is used to build cognitive transitions at the points where the structure is broken or the connection is interrupted. It is often manifested as auxiliary knowledge points, bridging concepts, or semantic intermediate nodes to ensure that the path from the starting node to the target node has cognitive coherence. It is controlled by the reasoning function. Based on the generation of graph semantic space, cognitive relationship pattern or task semantic similarity, it may be combined with graph neural network, knowledge embedding, language model and other implementation methods. is the transitional knowledge unit reasoning function, which indicates that the system generates reasonable The intelligent knowledge reasoning method called by the node is a response strategy function for the current graph state and potential energy difference. , determine whether to generate transitional knowledge units, how many to generate, and what content to generate. Its logic may include: querying semantically similar but unconnected nodes in the graph; automatically synthesizing knowledge points (such as inferring the distance judgment of positive and negative numbers on the number axis by comparing the "number axis" with "positive and negative numbers"); generating concept nodes based on inferring new vector center points in the semantic embedding vector space, It is the structural expansion amplification coefficient, which is an amplification parameter set to control the scale and granularity of structural repair. It indicates how many edge structures should be expanded per unit potential energy difference, and adjusts the repair "sensitivity" of the system when facing structural defects. : Careful repair, saving resources, suitable for minor structural drift; large : Radical repair, rapid restoration of large-scale fracture areas, It is the edge structure expansion, which indicates the number of new connecting edges that need to be added to the graph in order to restore the connectivity of the graph. Each edge represents a logical / cognitive relationship between two knowledge nodes in the graph, which is used to restore the connectivity loss caused by the interruption of the original path logic and to connect with the newly generated transitional knowledge unit set. Form an effective link, so that the graph structure is restored to a "pathway state" that can be used by the task scheduling module. It has a linear relationship and the amplification factor can be expanded by structure Amplify or compress the repair strength.

[0042] The inference function for transitional knowledge units refers to a function or model mechanism that, upon detecting structural defects in a knowledge graph (such as broken paths or isolated nodes), automatically generates new intermediary nodes (i.e., transitional knowledge units) based on the semantic features, structural relationships, and cognitive context of the missing region. Its core function is to fill cognitive gaps in the knowledge path, enabling students to smoothly transition from previously acquired knowledge to target knowledge points during task scheduling, avoiding sudden cognitive gaps. This function can be implemented based on existing intelligent inference models, such as graph neural networks (GCNs and GATs) for inferring highly connected nodes within the graph structure; knowledge embedding models (such as TransE and ComplEx) for identifying potential but missing relationship triples; or large-scale language models (such as GPT or BERT) for semantic interpolation, generating textual representations of knowledge fragments at structural gaps. This inference function not only serves as a logical generator for structural repair but also provides a key technical support for the graph's adaptive evolutionary capabilities, helping to achieve the leap from static explicit knowledge to dynamic implicit knowledge completion in knowledge graphs.

[0043] Through the above three-stage process, quantitative assessment of structural defects, identification of connection requirements and structural reconstruction are achieved, ensuring that the graph structure maintains cognitive rationality and task coherence during dynamic evolution.

[0044] Through continuous perception of multidimensional data and intelligent identification of structural defects during the operation of the knowledge graph, this invention can dynamically quantify the connectivity health of the graph. Based on the "knowledge flow potential difference" indicator, it locates areas of critical connectivity. It further infers and generates transitional knowledge units and implements edge structure expansion, thereby reconstructing a path structure that conforms to cognitive logic while maintaining semantic consistency. This mechanism effectively avoids task scheduling errors caused by graph disconnection, islands, or missing edges, improves the continuity and rationality of personalized task recommendations, and enhances the system's path generation stability and cognitive adaptability in real-world learning scenarios.

[0045] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0046] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0047] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0048] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0049] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0050] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0051] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0052] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0053] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0054] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. An intelligent learning and practice task generation algorithm for primary and secondary school students' knowledge mastery assessment based on knowledge graphs, characterized by: The following steps are involved: During the entire process of executing personalized learning and training task scheduling on the knowledge graph, multi-dimensional raw data is fully acquired, transforming the knowledge graph from a static structure to a dynamic observation operation state; Preprocess the acquired multi-dimensional raw data to obtain a data frame structure with unified dimensions; Using feature engineering methods, we extract key features that characterize the structural defects of the knowledge graph from the preprocessed data, and conduct a comprehensive analysis of the extracted key features to quantify the severity of the structural defects of the knowledge graph. The quantified key features are fed into a pre-trained machine learning model. The model identifies the patterns and distribution of structural anomalies and assesses in real time whether there are structural defects in the knowledge graph operation. When structural defects in the graph are identified, a "knowledge flow potential difference" indicator is constructed to measure the urgency of connection based on the semantic relevance of knowledge nodes, cognitive dependency order, and student behavioral characteristics. Guided by this indicator, transitional knowledge units are inferred and generated in the corresponding area, the edge structure is dynamically expanded, and new connection paths that conform to cognitive logic are constructed to adjust the local connectivity and structural layout of the graph.

2. The algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graph according to claim 1 is characterized in that: In the entire process of executing personalized learning and training task scheduling on the knowledge graph, in order to achieve the transformation from static structure to dynamic observation of running status, the specific steps of data acquisition are as follows: During the student-task interaction stage, the learning platform collects students’ response behavior data in real time during the question-answering process; In the task path scheduling phase, the graph scheduling engine collects node call records, path access status, and local connectivity indicator structure response data; During the model execution phase, the model execution log is generated through the scheduling module and the path planning algorithm interface, forming a system-level execution tracking chain; During the data collection stage, multi-dimensional data is aggregated in real time to the knowledge graph operation monitoring center through a unified scheduling and monitoring module.

3. The algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graph according to claim 1 is characterized in that: A feature engineering method is used to extract key features that characterize the structural defects of the knowledge graph from the preprocessed data. The extracted features include the proportion of cognitive spans exceeding a reasonable range in the task sequence and the path reconstruction structure change rate of the knowledge subgraph within the continuous task scheduling cycle. The proportion of cognitive spans exceeding a reasonable range in the task sequence and the path reconstruction structure change rate of the knowledge subgraph within the continuous task scheduling cycle are comprehensively analyzed under the detection window to generate cognitive logic fracture indicators and local structure drift indicators, respectively. The cognitive logic fracture indicators and local structure drift indicators are used to quantify the severity of the structural defects of the knowledge graph.

4. The algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graph according to claim 3 is characterized in that: The specific steps for comprehensively analyzing the proportion of cognitive spans exceeding the reasonable range in the task sequence under the detection window to generate cognitive logic fracture indicators are as follows: In the scheduling process of personalized chemistry learning tasks, the threshold of cognitive span is set to , which is used to limit the cognitive acceptance interval between adjacent task nodes. For any set of consecutive task nodes, the cognitive span difference in the knowledge graph is obtained, which is marked as ,when When , it is regarded as a cognitive rupture event. To characterize the intensity of each rupture event, the cognitive rupture intensity factor is calculated. The calculation expression is as follows: ， Where, It is The cognitive span difference between the group task node pairs, It is The spectral connection reachability marker factor of the group task node pair, It is The semantic embedding distance factor of the group task node pair, It is cognitive fracture intensity factor of group task node pairs; Under the detection window, suppose there are Group task node pairs, among which the ones that meet the break conditions are The cognitive jump events in all task paths were analyzed cumulatively to generate the cognitive logic fracture index. The calculation formula is as follows: ， Where, is the total number of task fragments that have broken in the detection window, is the total number of task fragments, that is, the total number of task node pairs, It is Path structure buffer parameters for non-broken segments, It is an indicator of cognitive logic rupture.

5. The algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graph according to claim 3 is characterized in that: The specific steps for comprehensively analyzing the path reconstruction structure change rate of the knowledge subgraph within the continuous task scheduling cycle under the detection window to generate the local structure drift index are as follows: In the detection window, select the target knowledge subgraph and record the task path sets generated in several consecutive task scheduling cycles. ,in Indicates that the target knowledge subgraph is in The set of valid task paths actually constructed within a scheduling cycle, is the number of scheduling cycles. To measure the degree of dynamic change of the graph path structure, a path structure change factor is introduced to measure the reconstruction amplitude of the path set between two adjacent scheduling cycles. The calculation expression of the path structure change factor is as follows: ， Where, is the path structure change factor, which indicates the difference between the two consecutive scheduling cycles. The scheduling cycle and The degree to which the task path structure of the target knowledge subgraph changes between scheduling cycles, The target knowledge subgraph is in The set of valid task paths actually constructed within a scheduling cycle, is the cardinality of the intersection of the path sets, is the cardinality of the union of the path sets; Based on the obtained structural change factors of all paths , the structural drift index is introduced to comprehensively characterize the volatility and intensity of structural changes. The local structural drift index is generated by weighted fusion of the difference amplitude and change intensity of the change factors in consecutive periods. The calculation expression is as follows: ， Where, is the natural base, is the trend amplification factor, It is The path structure change factor of the target knowledge subgraph between a scheduling cycle and the next cycle, is an indicator of local structural drift.

6. The algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graph according to claim 3 is characterized in that: The quantified cognitive logic fracture indicators and local structural drift indicators are input into a pre-trained machine learning model. The model generates a structural defect risk coefficient by identifying the pattern and distribution of structural abnormality features, and evaluates in real time whether there are structural defects in the operation of the knowledge graph based on the structural defect risk coefficient.

7. The algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graph according to claim 6 is characterized in that: The structural defect risk coefficient generated by the pre-trained machine learning model when identifying the pattern and distribution of structural abnormality features is compared with the pre-set structural defect risk coefficient reference threshold to evaluate in real time whether there are structural defects in the knowledge graph operation. The evaluation process is as follows: If the structural defect risk coefficient is greater than the pre-set structural defect risk coefficient reference threshold, it is judged that there is a structural defect in the operation of the knowledge graph; if the structural defect risk coefficient is less than or equal to the pre-set structural defect risk coefficient reference threshold, it is judged that there is no structural defect in the operation of the knowledge graph.

8. The algorithm for generating intelligent learning and practice tasks for evaluating knowledge mastery of primary and secondary school students based on knowledge graph according to claim 7 is characterized in that: After identifying structural defects in the graph, we construct a "knowledge flow potential difference" indicator to measure the urgency of connections based on the semantic relevance of knowledge nodes, cognitive dependency order, and student behavioral characteristics. Guided by this indicator, we infer and generate transitional knowledge units in the corresponding areas, dynamically expand the edge structure, and construct new connection paths that conform to cognitive logic. The specific steps are as follows: When structural defects are identified in the knowledge graph, a comprehensive evaluation is performed on the node characteristics of the defect-affected area to construct a regional weight factor that reflects the importance of the local structure and the urgency of repair. The calculation formula for the structural defect regional weight factor is as follows: ， Where, is the average semantic relatedness, is the average cognitive dependency order weight, is the average student behavior trigger frequency, The average semantic relatedness , average cognitive dependency order weight and average student behavioral trigger frequency The weight coefficient of is the structural defect area weight factor; Based on the obtained structural defect area weight factor , combined with the severity of structural defects, the "knowledge flow potential difference" indicator is calculated by the ratio of the current structural defect risk coefficient to the set reference threshold. It is used to characterize the structural tension and connectivity requirements brought about by the lack of connections in the current graph. The calculation expression of the "knowledge flow potential difference" indicator is as follows: ， Where, is the structural defect risk coefficient, is the reference threshold of the structural defect risk coefficient, It is the potential energy difference of knowledge flow; Knowledge flow potential difference To drive, inference is generated in the structural missing area, and the knowledge flow potential difference is generated. The numerical strength of dynamically expands the edge structure, builds new connection paths, reconstructs the cognitive chain, and realizes the adaptive repair of the local structure of the atlas. The formula is as follows: ， Where, It is a set of transitional knowledge units. is the transitional knowledge unit reasoning function, is the structural expansion magnification factor, is the edge structure extension.

Images