Knowledge network-oriented adaptive learning path planning method and system
By constructing a knowledge network and subgraph traversal, and combining mastery assessment to optimize the learning path, the problem of insufficient personalization of the learning path is solved, and learning efficiency and effectiveness are improved.
Patent Information
- Application Number
- CN202511835402.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-08
- Publication Date
- 2026-01-23
AI Technical Summary
Existing learning path planning cannot meet the individual needs of learners, resulting in low learning efficiency and effectiveness.
By constructing a knowledge network, setting target nodes, extracting multiple subgraphs, performing breadth-first traversal of path length and path congestion, and combining mastery assessment and feedback analysis, the learning path is optimized.
It enables dynamic adjustment of learning paths based on learners' individual needs, thereby improving learning efficiency and effectiveness.
Smart Images
Figure CN121388293A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of data processing, and particularly relates to a knowledge network-oriented adaptive learning path planning method and system. BACKGROUND
[0002] Traditional learning path planning methods often rely on teacher or expert experience. Teachers formulate unified learning plans and paths for learners according to teaching outlines or past teaching experience. For example, in school course learning, learning is carried out according to a predetermined course schedule. However, this learning path planning has obvious limitations and does not take into account individual differences of learners, such as different learning bases and learning abilities. Different learners have different knowledge learning and understanding abilities. For the same learning content, some learners may find it simple, while some learners may feel difficult and difficult to digest, thereby leading to poor learning efficiency and learning effect of learners and affecting learning interest and self-confidence of learners.
[0003] The prior art has the technical problem that learning path planning cannot meet the individualized level of learners, resulting in low learning efficiency and effect. SUMMARY
[0004] The purpose of the present application is to provide a knowledge network-oriented adaptive learning path planning method and system, which solves the technical problem that the prior art has the technical problem that learning path planning cannot meet the individualized level of learners, resulting in low learning efficiency and effect.
[0005] In view of the above problems, the present application provides a knowledge network-oriented adaptive learning path planning method and system.
[0006] In a first aspect, the present application provides a knowledge network-oriented adaptive learning path planning method, which comprises: constructing a knowledge network with knowledge points as nodes and the relationship from the first to the last knowledge point as directed edges; setting a target node and extracting a plurality of subgraphs that can reach the target node from the knowledge network; performing breadth-first traversal on the plurality of subgraphs from path length and path blocking degree to obtain an initial learning path; performing mastery degree evaluation on the initial learning path, and adaptively optimizing the initial learning path according to the mastery degree evaluation result to determine a target learning path.
[0007] Optionally, there is no loop in the knowledge network, and the relationship between any two nodes is direct connection or indirect connection through other nodes.
[0008] Optionally, the path length is analyzed by traversing the multiple subgraphs to obtain multiple path lengths; the path congestion is analyzed by traversing the multiple subgraphs to obtain multiple path congestion levels; and a breadth-first traversal is performed based on the multiple path lengths and the multiple path congestion levels to obtain the initial learning path.
[0009] Optionally, a first subgraph is extracted from the plurality of subgraphs, wherein the first subgraph includes M subgraph nodes, where M is a positive integer; big data mining of path blocking degree is performed on the M subgraph nodes respectively to obtain a set of path blocking degree of the M subgraph nodes; directional filtering is performed on the set of path blocking degree of the M subgraph nodes respectively to obtain a set of filtered path blocking degree of the M subgraph nodes; based on the first subgraph, the filtered path blocking degree of the M subgraph nodes is weighted to determine a first path blocking degree, and the first path blocking degree is added to the plurality of path blocking degree.
[0010] Optionally, M initial directional subgraph node path blocking degrees are extracted from the set of M subgraph node path blocking degrees to construct M initial directional subgraph node path blocking degree screening neighborhoods; within the M initial directional subgraph node path blocking degree screening neighborhoods, directional screening is performed on the M initial directional subgraph node path blocking degrees respectively to determine M iterative directional subgraph node path blocking degrees and M iterative directional subgraph node path blocking degree screening neighborhoods; it is determined whether the neighborhood density coefficient of the M iterative directional subgraph node path blocking degree screening neighborhoods is greater than or equal to the M initial directional subgraph node path blocking degree screening neighborhoods. If so, within the M iterative directional subgraph node path blocking degree screening neighborhoods, directional screening is performed on the M iterative directional subgraph node path blocking degrees respectively until the neighborhood density coefficient of the currently obtained directional subgraph node path blocking degree screening neighborhood is less than the neighborhood density coefficient of the previously obtained directional subgraph node path blocking degree screening neighborhood.
[0011] Optionally, if the neighborhood density coefficient of the M iterative directional subgraph node path blocking degree screening neighborhood is less than the M initial directional subgraph node path blocking degree screening neighborhood, then the M initial directional subgraph node path blocking degree is used as the M subgraph node screening path blocking degree.
[0012] Optionally, knowledge point resources are pushed based on the initial learning path, and the push feedback results are captured; the mastery assessment tool is called to analyze the initial learning path and the push feedback results to obtain the mastery assessment results.
[0013] Optionally, multiple sample learning paths and multiple sample push feedback results, as well as the corresponding multiple sample mastery evaluation results, are obtained as training data; the framework built on the feedforward neural network is trained using the training data until convergence, thereby obtaining the mastery evaluator.
[0014] Optionally, a preset feedback optimization window is obtained; the path blocking degree of the target learning path is analyzed in the preset feedback optimization window, and learning content is pushed according to the analysis results.
[0015] A second aspect of this application provides an adaptive learning path planning system for knowledge networks. The system includes: a knowledge network construction module for constructing a knowledge network using knowledge points as nodes and directed edges representing the sequential relationships between knowledge points; a subgraph extraction module for setting a target node and extracting multiple subgraphs from the knowledge network that can reach the target node; an initial learning path acquisition module for performing a breadth-first traversal of the multiple subgraphs based on path length and path congestion level to obtain an initial learning path; and a target learning path determination module for evaluating the mastery level of the initial learning path and adaptively optimizing the initial learning path based on the mastery level evaluation results to determine the target learning path.
[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: The method provided in this application constructs a knowledge network using knowledge points as nodes and directed edges representing the sequential relationships between knowledge points. It then sets a target node and extracts multiple subgraphs from the knowledge network that can reach the target node. A breadth-first traversal of these subgraphs is performed based on path length and path congestion to obtain an initial learning path. The mastery level of the initial learning path is evaluated, and adaptive optimization is performed based on the evaluation results to determine the target learning path. This achieves the technical effect of dynamically adjusting the learning path according to the learner's personalized needs, thereby improving learning efficiency and effectiveness.
[0017] The above description is merely an overview of the technical solution of this application. To enable a clearer understanding of the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0019] Figure 1 A flowchart illustrating the adaptive learning path planning method for knowledge networks provided in this application.
[0020] Figure 2 A schematic diagram of the structure of the adaptive learning path planning system for knowledge networks provided in this application.
[0021] Figure labeling: Knowledge network construction module 11, subgraph extraction module 12, initial learning path acquisition module 13, target learning path determination module 14. Detailed Implementation
[0022] This application provides an adaptive learning path planning method and system for knowledge networks, addressing the technical problem that existing technologies fail to meet the individualized learning levels of learners, resulting in low learning efficiency and effectiveness. It achieves the technical effect of dynamically adjusting the learning path according to the learner's individual needs, thereby improving learning efficiency and effectiveness.
[0023] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. It should also be noted that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, not all of them.
[0024] Example 1 like Figure 1 As shown, this application provides an adaptive learning path planning method for knowledge networks, which includes: A knowledge network is constructed using knowledge points as nodes and the relationships between knowledge points from beginning to end as directed edges.
[0025] Specifically, knowledge content within the target learning domain is collected from various channels, such as textbook content, course outlines, and expert experience. This knowledge content is then analyzed, using techniques like NLP (Natural Language Processing) to identify technical terms and keywords from the text through algorithms such as named entity recognition and keyword extraction, thus obtaining knowledge points within the target learning domain. These extracted knowledge points are then processed according to subject, theme, or level, representing them as uniquely identified node data. Each node is linked to the corresponding explanation and practice data for that knowledge point. Through expert experience, course outlines, or textbook content, the dependencies between knowledge points are analyzed, determining which knowledge points are foundational or prerequisites for others, and clarifying the sequential order of knowledge points. For example, when learning advanced mathematics, one must first understand fundamental concepts such as functions and limits before understanding the concepts of derivatives and integrals.
[0026] A knowledge network is represented using directed graphs from graph theory, where knowledge points are nodes and the order of knowledge points is represented by directed edges. The connections between knowledge points are described using a unidirectional directed graph, thus constructing the knowledge network. Each node in the knowledge network corresponds to one knowledge point, and each edge reflects the order of two knowledge points.
[0027] By constructing a knowledge network, scattered knowledge content is integrated into a whole, forming a clear knowledge framework. This allows learners to intuitively see the connections and hierarchical relationships between various knowledge points. Furthermore, the edges of the knowledge network can be used to plan learning paths, effectively improving the rationality and adaptability of learning path planning.
[0028] Furthermore, there are no loops within the knowledge network, and the relationship between any two nodes is either a direct connection or an indirect connection through other nodes.
[0029] Specifically, the knowledge network contains no cycles. That is, in the directed graph constructed by the knowledge network, there is no path that starts from a knowledge point, traverses several directed edges, and returns to the same knowledge point. The learning order of knowledge points is unidirectional and non-cyclic, avoiding circular dependencies or disordered learning order in the learning path. Furthermore, the relationship between any two nodes in the knowledge network is either a direct connection or an indirect connection through other nodes. A direct connection means that two knowledge points are directly connected by a directed edge, while an indirect connection means that two knowledge points are not directly connected by a directed edge, but are connected through other knowledge points and directed edges.
[0030] Direct connections clarify the dependencies between knowledge points, allowing learners to understand that understanding the prerequisite knowledge points directly related to a particular knowledge point is essential for learning it. Indirect connections reflect the breadth and interconnectedness of the knowledge network. By analyzing both direct and indirect connections between knowledge points, the scientific accuracy and effectiveness of learning path planning can be improved, achieving a smooth transition between learning paths and a hierarchical progression of knowledge.
[0031] Set a target node, and extract multiple subgraphs from the knowledge network that can reach the target node.
[0032] Specifically, a target node is defined, which represents the knowledge content the learner aims to acquire, such as a course unit. The directed edges in the knowledge network are reversed to form a reverse graph. A breadth-first search or depth-first search is used to traverse the reverse graph from the target node outwards, expanding layer by layer towards its predecessor nodes, recording each visited edge and node until the node is no longer expandable, thus obtaining multiple paths originating from the target node. These paths are then reorganized according to the original knowledge network direction, forming multiple subgraphs that can reach the target node from the knowledge network. Each subgraph reflects the path from various prerequisite knowledge points to the target node.
[0033] The extracted subgraph expands based on the target node, which can accurately locate knowledge points closely related to the target node, thereby improving the relevance and effectiveness of learning.
[0034] The initial learning path is obtained by performing a breadth-first traversal on the multiple subgraphs based on path length and path congestion level.
[0035] Furthermore, based on path length and path congestion level, a breadth-first traversal is performed on the multiple subgraphs to obtain an initial learning path, including: traversing the multiple subgraphs to analyze path length and obtain multiple path lengths; traversing the multiple subgraphs to analyze path congestion level and obtain multiple path congestion levels; and performing a breadth-first traversal based on the multiple path lengths and the multiple path congestion levels to obtain the initial learning path.
[0036] Specifically, in the subgraphs of the knowledge network, path length refers to the number of nodes traversed from the starting knowledge point to the target node, reflecting the hierarchical depth and required knowledge span of the learning path. The starting node is the most basic knowledge point in the subgraph related to the target node. The path lengths of multiple subgraphs are analyzed by traversing them, calculating the path length for each subgraph with each node as the starting point and the target node as the ending point. Simultaneously, path congestion analysis is performed by traversing multiple subgraphs. Path congestion refers to the number of knowledge points in the path that are difficult to understand or require additional learning costs. These knowledge points are the difficult, obscure, or less related knowledge points in the knowledge network, representing the difficulty level of the learning path. A congestion score can be assigned to each node, determined based on the difficulty of the knowledge point, the learner's prior knowledge, and the strength of the connections between knowledge points.
[0037] Path congestion analysis is introduced during breadth-first traversal of paths. Weights are assigned to both path length and path congestion level, and these weights can be dynamically adjusted based on learners' knowledge level and learning goals. For example, for beginners, the weight of path congestion level can be increased to prioritize paths with lower congestion levels, reducing learning difficulty. For learners with some experience, the weight of path length can be increased to select shorter paths and improve learning efficiency. Multiple path lengths and congestion levels are standardized, and then weighted and summed according to the assigned weights to obtain a comprehensive score for each path. The formula is: Comprehensive Score = Path Length Weight × Path Length + Path Congestion Level Weight × Path Congestion Level. For example, if the path length weight is 0.4, the path congestion level weight is 0.6, and a path length of 5 with a congestion level of 3 in a subgraph, then the comprehensive score for that path is 0.4 × 5 + 0.6 × 3 = 3.8. After completing the comprehensive analysis of multiple subgraph paths, the paths are sorted in order of score from low to high, and the path with the lowest comprehensive score is selected as the current initial learning path. The initial learning path includes a series of knowledge points in a learning order.
[0038] By considering both path length and path congestion, a relatively short and easy-to-understand path can be selected as the initial learning path. This allows learners to learn the target knowledge more efficiently within a limited learning time, avoiding the inefficiency caused by choosing overly long or complex paths.
[0039] Furthermore, the path congestion level is analyzed by traversing the multiple subgraphs to obtain multiple path congestion levels, including: extracting a first subgraph from the multiple subgraphs, wherein the first subgraph includes M subgraph nodes, where M is a positive integer; performing big data mining on the path congestion level of each of the M subgraph nodes to obtain a set of path congestion levels of the M subgraph nodes; performing directional filtering on the set of path congestion levels of the M subgraph nodes to obtain a set of filtered path congestion levels of the M subgraph nodes; and, based on the first subgraph, weighting the filtered path congestion levels of the M subgraph nodes to determine a first path congestion level, and adding the first path congestion level to the multiple path congestion levels.
[0040] Specifically, a first subgraph is extracted from multiple subgraphs. "First" does not represent any particular order but refers to any one of the multiple subgraphs. Each subgraph contains a subset of nodes and relationships within the target learning domain. The first subgraph comprises M subgraph nodes, where M is a positive integer. Subgraph nodes refer to specific knowledge points within the knowledge network. For each of the M extracted subgraph nodes, path blocking is analyzed using big data mining. This includes analyzing the degree to which each node becomes a path blocking initiation point and the degree to which each node serves as a connection point. The degree to which each node becomes a path blocking initiation point refers to the extent to which a node, when used as the starting point of a learning path, might obstruct the learning path due to factors such as the complexity and abstractness of its knowledge points. For example, by collecting a large amount of learner data, the frequency with which learners encounter difficulties, stagnate, or require additional learning materials when starting from a particular node during the learning process can be analyzed, quantifying the degree to which that node becomes a path blocking initiation point. The degree to which each node acts as a link refers to the extent to which a node is connected and supported by other nodes in a knowledge network. If a node has complex, multi-linked relationships with many other nodes, learners need to master multiple related pieces of knowledge simultaneously to understand that node, increasing the obstruction of the learning path. For example, by analyzing the in-degree of nodes and the association weights between nodes in a knowledge network, we can analyze the degree to which each node acts as a link. The in-degree is the number of edges pointing to a node, reflecting the degree to which that node receives information in the knowledge network, i.e., there are multiple other nodes that have knowledge transfer input relationships with that node. The association weights between nodes reflect the tightness of knowledge transfer or connection between two nodes. Experts in relevant fields can assign weights to the edges between nodes based on the importance and relevance of the knowledge. Through big data mining of path obstruction, we obtain a set of path obstruction degrees for M subgraph nodes. Each element in the set corresponds to relevant information about the path obstruction degree of a graph node.
[0041] Directional filtering is performed on the path blocking degree sets of M subgraph nodes. Directional filtering refers to further refining the assessment of path blocking degree based on the directed relationships between nodes in the knowledge network. For example, in a knowledge network, if node A points to node B, it indicates a knowledge transfer relationship between node A and node B. If, when analyzing the path blocking degree of node B, key difficulties are found in the knowledge transfer path from node A to node B, then the path blocking degree of node B may be affected by node A. By analyzing the edge weights and characteristics of nodes on the path in the knowledge network, the data in the subgraph node path blocking degree sets are filtered and adjusted to obtain the path blocking degree of M subgraph nodes.
[0042] The path blocking degree of the M subgraph nodes in the first subgraph is weighted and analyzed to comprehensively assess the importance and difficulty of each node in the first subgraph, thus determining the overall path blocking degree of the first subgraph, which is then added to multiple path blocking degree sets. For example, if a node is closely related to several other nodes in the first subgraph, its path blocking degree has a significant impact on the overall path blocking degree of the first subgraph, and its weight can be set higher. The weight of each node in the first subgraph is determined by analyzing node centrality indicators such as degree centrality, proximity centrality, and betweenness centrality. The path blocking degrees of the M subgraph nodes are then weighted and summed based on these weights to obtain the first path blocking degree.
[0043] By analyzing the path congestion levels of multiple subgraphs, we can obtain the learning difficulty of each node in each subgraph and the subgraph as a whole. This provides learners with more precise learning path planning, avoiding excessive difficulties and obstacles during the learning process and improving learning efficiency and quality. Furthermore, path congestion analysis can also provide important references for knowledge instruction and curriculum design, helping teachers better organize teaching content, rationally arrange teaching progress, and improve teaching quality and effectiveness.
[0044] Furthermore, directional filtering is performed on the set of path blocking degrees for the M subgraph nodes to obtain M selected path blocking degrees for the subgraph nodes. This includes: extracting M initial directional subgraph node path blocking degrees from the set of M subgraph node path blocking degrees, constructing a filtering neighborhood for the M initial directional subgraph node path blocking degrees; and performing directional filtering on the M initial directional subgraph node path blocking degrees within the filtering neighborhood for the M initial directional subgraph node path blocking degrees to determine M iterative directional subgraph node path blocking degrees and M iterative directional subgraph nodes. The path blocking degree of the node is filtered into neighborhoods. It is determined whether the neighborhood density coefficient of the path blocking degree filtering neighborhoods of the M iterative directional subgraph nodes is greater than or equal to the path blocking degree filtering neighborhoods of the M initial directional subgraph nodes. If so, the path blocking degree of the M iterative directional subgraph nodes is filtered in a directional manner within the path blocking degree filtering neighborhoods of the M iterative directional subgraph nodes, until the neighborhood density coefficient of the path blocking degree filtering neighborhoods of the current directional subgraph nodes is less than the neighborhood density coefficient of the path blocking degree filtering neighborhoods of the previous directional subgraph nodes.
[0045] Specifically, in a knowledge network, nodes have directed dependencies. M initial directional subgraph node path blocking degrees are extracted from a set of M subgraph node path blocking degrees. These initial directional subgraph node path blocking degrees refer to the blocking difficulty value distributed along the learning path direction, initially calculated for each subgraph node based on its directed dependencies within the knowledge network. Using each initial directional subgraph node path blocking degree as a center, and based on its position in the knowledge network space and the dependency strength between adjacent nodes, M initial directional subgraph node path blocking degree filtering neighborhoods are constructed. These M initial directional subgraph node path blocking degree filtering neighborhoods represent the set of nodes that have a direct or indirect dependency on the learning path with the node corresponding to the initial directional subgraph node path blocking degree. For example, using a graph traversal algorithm, starting from the center node, adjacent nodes are traversed according to a set distance metric, such as the shortest path length between nodes in the knowledge network. All relevant node path blocking degree data that satisfy the distance metric condition are integrated to form the initial directional subgraph node path blocking degree filtering neighborhoods.
[0046] Within the initial directional subgraph node path congestion degree filtering neighborhood, directional filtering is performed on the M initial directional subgraph node path congestion degrees. For example, a mean-shift algorithm is used for directional filtering. The mean-shift algorithm is a non-parametric clustering algorithm based on density gradient ascent, which obtains dense regions of data by iteratively finding the mean shift direction of data points. For each of the M initial directional subgraph node path congestion degree filtering neighborhoods, using the data points within the initial directional subgraph node path congestion degree filtering neighborhood as samples, the mean of multiple data points in the subgraph node path congestion degree set is calculated. The center point, i.e., the initial directional subgraph node path congestion degree, is moved a certain distance in the mean direction to obtain a new center point, i.e., the M iterative directional subgraph node path congestion degrees. At the same time, the range of the filtering neighborhood is redefined based on the new center point, resulting in the M iterative directional subgraph node path congestion degree filtering neighborhoods. For example, in the set of path blocking degrees of subgraph nodes, the initial center point is C0. The mean value of the data points in the initial directional subgraph node path blocking degree screening neighborhood is calculated as M0. C0 is moved to M0 by a certain step to obtain C1. With C1 as the center, the iterative directional subgraph node path blocking degree screening neighborhood is obtained according to the distance metric rule.
[0047] Calculate the neighborhood density coefficient of the filtering neighborhood for the path blocking degree of M iterative directional subgraph nodes. The neighborhood density coefficient is the ratio of the number of nodes in the filtering neighborhood to the number of nodes in the corresponding subgraph node path blocking degree set. The larger the ratio, the more concentrated the data in the neighborhood, and the more representative the corresponding path blocking degree assessment. For example, if the initial filtering neighborhood for the path blocking degree of the directional subgraph nodes contains 2 data nodes, and the corresponding overall subgraph node path blocking degree set contains 10 data points, then the neighborhood density coefficient of the initial filtering neighborhood for the path blocking degree of the directional subgraph nodes is 0.2. The neighborhood density coefficients of the M iterative directional subgraph node path blocking degree screening neighborhoods are compared with the M initial directional subgraph node path blocking degree screening neighborhoods. If the neighborhood density coefficient of the M iterative directional subgraph node path blocking degree screening neighborhoods is greater than or equal to that of the M initial directional subgraph node path blocking degree screening neighborhoods, then within the M iterative directional subgraph node path blocking degree screening neighborhoods, directional screening is performed on the path blocking degree of the M iterative directional subgraph node paths, and the mean drift algorithm is executed iteratively. The above process of calculating the mean, moving the center point, and determining the new neighborhood is repeated until the neighborhood density coefficient of the newly obtained directional subgraph node path blocking degree screening neighborhood is less than that of the previously obtained directional subgraph node path blocking degree screening neighborhood. This indicates that the current neighborhood has reached a relatively optimal density state, and the corresponding subgraph node path blocking degree is taken as the M subgraph node path blocking degree.
[0048] By performing directional filtering on the path degree set of subgraph nodes, the correlation between nodes was analyzed, enabling a more accurate analysis of the path blocking situation of each subgraph node in different knowledge transfer directions. This provides more reliable, accurate, and comprehensive data support for learning path planning and knowledge teaching optimization, improving the effectiveness and applicability of path planning and enhancing learning efficiency.
[0049] Furthermore, if the neighborhood density coefficient of the filtering neighborhood of the M iterative directional subgraph node path blocking degree is less than that of the filtering neighborhood of the M initial directional subgraph node path blocking degree, then the path blocking degree of the M initial directional subgraph node is used as the path blocking degree of the M subgraph node filtering.
[0050] Specifically, when the neighborhood density coefficient of the neighborhood selected based on the path blocking degree of the M iterative directional subgraph nodes is less than that of the neighborhood selected based on the path blocking degree of the M initial directional subgraph nodes, it indicates that the iteration process has not made the data of the neighborhood selected based on the path blocking degree of the M initial directional subgraph nodes more concentrated; instead, it shows a trend of dispersion. Therefore, directly using the path blocking degree of the M initial directional subgraph nodes as the path blocking degree of the M subgraph nodes avoids errors caused by iteration, ensures the reliability of the path blocking degree of the M subgraph nodes, and thus improves the reliability and effectiveness of adaptive learning path planning, thereby improving the learner's learning efficiency.
[0051] The initial learning path is assessed for mastery, and the initial learning path is adaptively optimized based on the mastery assessment results to determine the target learning path.
[0052] Furthermore, the initial learning path is assessed for mastery, and the initial learning path is adaptively optimized based on the mastery assessment results to determine the target learning path. This includes: pushing knowledge point resources based on the initial learning path and capturing the push feedback results; and calling the mastery assessor to analyze the initial learning path and the push feedback results to obtain the mastery assessment results.
[0053] Specifically, based on the initial learning path obtained, knowledge point materials are pushed to learners. When pushing these materials, the system sequentially selects and pushes relevant learning materials for each knowledge point from a rich knowledge resource library, such as instructional videos, documents, and online exercises, according to the initial learning path. After pushing resources, feedback results are collected in real time. These feedback results include learners' various reactions to each knowledge point's resources, including the time spent viewing the materials and the accuracy rate of completing the exercises. Push feedback data can be obtained by setting up relevant tracking mechanisms on the learning platform, such as embedding tracking code in web pages to record learners' click behavior and listening time; for online exercises, it can obtain learners' answer performance and accuracy rate.
[0054] After receiving the push feedback results, a mastery assessment tool is invoked to comprehensively analyze the initial learning path and the push feedback results, obtaining a mastery assessment result. This mastery assessment tool, built on machine learning technology, receives and analyzes data from the initial learning path and push feedback results, outputting a mastery assessment result, which is a score or category label representing the learner's mastery of each knowledge point in the initial learning path. The initial learning path is adaptively optimized based on the mastery assessment results. If the assessment results show that the learner has a low mastery of certain knowledge points, the path for those knowledge points is optimized, providing more basic and related knowledge points. Conversely, if the learner has a high mastery of certain knowledge points, the learning of those knowledge points can be simplified, reducing repetitive learning resource pushes and accelerating the learning progress. The order of knowledge points in the initial learning path is adaptively adjusted based on the mastery assessment results to obtain the target learning path.
[0055] By pushing knowledge point resources and capturing feedback results based on the initial learning path, learners' learning status and needs can be understood in real time. By calling the mastery assessment tool for analysis, learners' mastery of each knowledge point can be accurately quantified. Based on the mastery results, the initial learning path can be adaptively optimized to ensure that the target learning path always matches the learner's actual level and learning progress, thereby improving learning efficiency and effectiveness, and enhancing learners' learning experience and satisfaction.
[0056] Furthermore, before calling the mastery evaluator, the process includes: obtaining multiple sample learning paths and multiple sample push feedback results, as well as the corresponding multiple sample mastery evaluation results as training data; using the training data to train the framework built on the feedforward neural network until convergence, thereby obtaining the mastery evaluator.
[0057] Specifically, a large amount of sample data is collected from actual learning scenarios using knowledge networks, including multiple sample learning paths, multiple sample push feedback results, and corresponding multiple sample mastery assessment results. The sample learning path refers to the order in which different learners learn knowledge points under a specific learning objective, which can be obtained through learners' historical learning records or learning calculations. The sample push feedback results reflect learners' responses to the knowledge point materials pushed to them according to the sample learning path. The sample mastery assessment results are a quantitative or categorical evaluation of learners' mastery of knowledge points under the corresponding learning path, for example, represented by scores of 0-100, or by categories such as not mastered, basically mastered, and proficiently mastered, evaluated by teachers or domain experts based on learners' practice and exam performance. Multiple sample learning paths, multiple sample push feedback results, and corresponding multiple sample mastery assessment results are used as training data. A mastery assessment framework is constructed based on a feedforward neural network, where the neural network input layer comprehensively reflects the characteristics of learners' learning paths and push feedback information, including features such as learning paths and push feedback results. For example, path length, path congestion level in sample learning paths, and various indicators in sample push feedback results, such as click-through rate, dwell time, and practice accuracy, are used as input features. The output layer outputs the corresponding mastery assessment value, which is the learner's level of mastery of the knowledge points.
[0058] Training data is input into the neural network for training. Through training, the weights and biases of the neural network are continuously adjusted to ensure that the output evaluation value is as accurate as possible in matching the actual mastery assessment result. During training, a backpropagation algorithm is used to calculate the error between the output result h and the actual sample mastery assessment result, and this error is backpropagated into the neural network to optimize the network parameters. Convergence is achieved by minimizing the loss function, such as mean squared error. After training, the resulting mastery evaluator can assess the learner's mastery of each knowledge point in real time based on the input learning path and feedback data. By continuously optimizing the learning path, it ensures that the learning content matches the student's cognitive progress, thereby improving the adaptability and effectiveness of learning path planning, increasing learner learning efficiency, and making learning path planning more scientific and personalized.
[0059] Furthermore, the mastery assessment of the initial learning path is performed, and the initial learning path is adaptively optimized based on the mastery assessment results to determine the target learning path. This also includes: obtaining a preset feedback optimization window; performing path obstruction analysis on the target learning path within the preset feedback optimization window; and pushing learning content based on the analysis results.
[0060] Specifically, a preset feedback optimization window is obtained. This preset feedback optimization window is a pre-defined time range or learning progress interval, providing clear time or progress boundaries for path congestion analysis and learning content delivery. For example, the target learning course is divided into several knowledge points according to the learning path, and then divided according to the number of knowledge points learned, with each window consisting of three completed knowledge points.
[0061] After entering the preset feedback optimization window, a path blocking analysis is performed on the target learning path. Path blocking refers to a lack of mastery of a particular knowledge point at a given node; if one node is blocked, all nodes reachable from that node are also blocked. Path blocking can be calculated by collecting various learning behavior data from the learner within the preset feedback optimization window. For example, analyzing the learner's error rate when completing exercises or assignments; if the error rate consistently exceeds a certain threshold, it indicates that the learner has difficulty learning that knowledge point, and the path blocking is high. Learning content is then pushed based on the analysis results. If the analysis shows a high path blocking for a particular knowledge point, it indicates that the learner has not mastered that knowledge point well, and learning resources related to that knowledge point are pushed. During the push process, an intelligent recommendation algorithm can be used, based on collaborative filtering, to find the learning records of other learners with similar learning difficulties and recommend learning content that can help solve similar problems. For example, if it is found that many learners at a similar level have mastered the same function definition problem by watching a specific instructional video, that video is recommended to the current learner. At the same time, different types of learning content can be pushed according to different levels of path congestion. For cases with low congestion, some extended exercises or simple review materials can be pushed, while for cases with high congestion, more detailed teaching explanations, case studies, etc. can be provided.
[0062] By providing a pre-defined feedback optimization window, a clear time and progress framework is offered to learners, improving the effectiveness and timeliness of the analysis. Path obstruction analysis of the target learning path accurately identifies the difficulties learners encounter, improving the accuracy of personalized recommendations. This, in turn, enhances the reliability and relevance of learning content delivery, ultimately improving learner outcomes.
[0063] Example 2 Based on the same inventive concept as the adaptive learning path planning method for knowledge networks in the foregoing embodiments, such as Figure 2 As shown, this application provides an adaptive learning path planning system for knowledge networks, wherein the adaptive learning path planning system for knowledge networks includes: The knowledge network construction module 11 is used to construct a knowledge network with knowledge points as nodes and the sequential relationship between knowledge points as directed edges; the subgraph extraction module 12 is used to set a target node and extract multiple subgraphs from the knowledge network that can reach the target node; the initial learning path acquisition module 13 is used to perform a breadth-first traversal of the multiple subgraphs based on path length and path blocking degree to obtain an initial learning path; the target learning path determination module 14 is used to evaluate the mastery of the initial learning path and adaptively optimize the initial learning path based on the mastery evaluation results to determine the target learning path.
[0064] Furthermore, the knowledge network construction module 11 also includes: there are no loops in the knowledge network, and the relationship between any two nodes is a direct connection or an indirect connection through other nodes.
[0065] Furthermore, the initial learning path acquisition module 13 further includes: traversing the multiple subgraphs to perform path length analysis and obtain multiple path lengths; traversing the multiple subgraphs to perform path blocking degree analysis and obtain multiple path blocking degrees; and performing breadth-first traversal based on the multiple path lengths and the multiple path blocking degrees to obtain the initial learning path.
[0066] Furthermore, the initial learning path acquisition module 13 further includes: extracting a first subgraph from the plurality of subgraphs, wherein the first subgraph includes M subgraph nodes, where M is a positive integer; performing big data mining on the path blocking degree of each of the M subgraph nodes to obtain a set of path blocking degrees of the M subgraph nodes; performing directional filtering on the set of path blocking degrees of the M subgraph nodes to obtain a set of filtered path blocking degrees of the M subgraph nodes; and, based on the first subgraph, weighting the filtered path blocking degrees of the M subgraph nodes to determine a first path blocking degree, and adding the first path blocking degree to the plurality of path blocking degrees.
[0067] Furthermore, the initial learning path acquisition module 13 further includes: extracting M initial directional subgraph node path blocking degrees from the set of M subgraph node path blocking degrees, and constructing M initial directional subgraph node path blocking degree screening neighborhoods; within the M initial directional subgraph node path blocking degree screening neighborhoods, performing directional screening on the M initial directional subgraph node path blocking degrees respectively, to determine M iterative directional subgraph node path blocking degrees and M iterative directional subgraph node path blocking degree screening neighborhoods; determining whether the neighborhood density coefficient of the M iterative directional subgraph node path blocking degree screening neighborhoods is greater than or equal to the M initial directional subgraph node path blocking degree screening neighborhoods respectively, and if so, then within the M iterative directional subgraph node path blocking degree screening neighborhoods, performing directional screening on the M iterative directional subgraph node path blocking degrees respectively, until the neighborhood density coefficient of the currently obtained directional subgraph node path blocking degree screening neighborhood is less than the neighborhood density coefficient of the previously obtained directional subgraph node path blocking degree screening neighborhood.
[0068] Furthermore, the initial learning path acquisition module 13 further includes: if the neighborhood density coefficient of the M iterative directional subgraph node path blocking degree screening neighborhood is less than the M initial directional subgraph node path blocking degree screening neighborhood, then the M initial directional subgraph node path blocking degree is used as the M subgraph node screening path blocking degree.
[0069] Furthermore, the target learning path determination module 14 also includes: pushing knowledge point resources based on the initial learning path and capturing the push feedback results; calling the mastery assessment tool to analyze the initial learning path and the push feedback results to obtain mastery assessment results.
[0070] Furthermore, the target learning path determination module 14 also includes: acquiring multiple sample learning paths and multiple sample push feedback results, as well as the corresponding multiple sample mastery evaluation results as training data; using the training data to train the framework built based on the feedforward neural network until convergence, thereby obtaining the mastery evaluator.
[0071] Furthermore, the target learning path determination module 14 also includes: obtaining a preset feedback optimization window; performing path obstruction analysis on the target learning path in the preset feedback optimization window, and pushing learning content based on the analysis results.
[0072] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The adaptive learning path planning method and specific examples for knowledge networks in the aforementioned Embodiment 1 are also applicable to the adaptive learning path planning system for knowledge networks in this embodiment. Through the foregoing detailed description of the adaptive learning path planning method for knowledge networks, those skilled in the art can clearly understand the adaptive learning path planning system for knowledge networks in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.
[0073] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
[0074] Obviously, those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.
Claims
1. An adaptive learning path planning method for knowledge networks, characterized in that, The method includes: A knowledge network is constructed using knowledge points as nodes and the relationships between knowledge points from beginning to end as directed edges. Set a target node, and extract multiple subgraphs from the knowledge network that can reach the target node; Based on path length and path congestion level, a breadth-first traversal is performed on the multiple subgraphs to obtain the initial learning path; The initial learning path is assessed for mastery, and the initial learning path is adaptively optimized based on the mastery assessment results to determine the target learning path.
2. The adaptive learning path planning method for knowledge networks as described in claim 1, characterized in that, There are no loops in the knowledge network, and the relationship between any two nodes is either a direct connection or an indirect connection through other nodes.
3. The adaptive learning path planning method for knowledge networks as described in claim 1, characterized in that, Based on path length and path congestion level, a breadth-first traversal is performed on the multiple subgraphs to obtain initial learning paths, including: Traverse the multiple subgraphs to perform path length analysis and obtain multiple path lengths; The path blocking degree is analyzed by traversing the multiple subgraphs to obtain the blocking degree of multiple paths. The initial learning path is obtained by performing a breadth-first traversal based on the lengths and congestion levels of the multiple paths.
4. The adaptive learning path planning method for knowledge networks as described in claim 3, characterized in that, The path congestion level is analyzed by traversing the multiple subgraphs to obtain multiple path congestion levels, including: Extract a first subgraph from the plurality of subgraphs, wherein the first subgraph comprises M subgraph nodes, where M is a positive integer; Perform big data mining on the path blocking degree of each of the M subgraph nodes to obtain a set of path blocking degree of the M subgraph nodes; The path blocking degree sets of the M subgraph nodes are filtered in a directional manner to obtain the path blocking degree of the M subgraph nodes. Based on the first subgraph, the path blocking degree of the M subgraph nodes is weighted to determine the first path blocking degree, and the first path blocking degree is added to the multiple path blocking degree.
5. The adaptive learning path planning method for knowledge networks as described in claim 4, characterized in that, The path blocking degree sets of the M subgraph nodes are filtered in a directional manner to obtain the path blocking degree of the M subgraph nodes, including: From the set of path blocking degrees of the M subgraph nodes, M initial directional subgraph node path blocking degrees are extracted, and M initial directional subgraph node path blocking degree filtering neighborhoods are constructed. Within the filtering neighborhood of the path blocking degree of the M initial directional subgraph nodes, the path blocking degree of the M initial directional subgraph nodes is filtered in a directional manner to determine the path blocking degree of the M iterative directional subgraph nodes and the filtering neighborhood of the path blocking degree of the M iterative directional subgraph nodes. Each of the M iterative directional subgraph node path blocking degree screening neighborhoods is determined to be greater than or equal to the M initial directional subgraph node path blocking degree screening neighborhoods. If so, then within the M iterative directional subgraph node path blocking degree screening neighborhoods, the path blocking degree of the M iterative directional subgraph node is directionally screened until the neighborhood density coefficient of the current directional subgraph node path blocking degree screening neighborhood is less than the neighborhood density coefficient of the previous directional subgraph node path blocking degree screening neighborhood.
6. The adaptive learning path planning method for knowledge networks as described in claim 5, characterized in that, If the neighborhood density coefficient of the filtering neighborhood for the path blocking degree of the M iterative directional subgraph nodes is less than that of the filtering neighborhood for the path blocking degree of the M initial directional subgraph nodes, then the path blocking degree of the M initial directional subgraph nodes is used as the path blocking degree of the filtering neighborhood for the M subgraph nodes.
7. The adaptive learning path planning method for knowledge networks as described in claim 1, characterized in that, The initial learning path is assessed for mastery, and the initial learning path is adaptively optimized based on the mastery assessment results to determine the target learning path, including: Based on the initial learning path, knowledge point resources are pushed out, and the push feedback results are captured. The mastery assessment tool is invoked to analyze the initial learning path and the push feedback results to obtain the mastery assessment results.
8. The adaptive learning path planning method for knowledge networks as described in claim 7, characterized in that, include: The training data includes multiple sample learning paths, multiple sample push feedback results, and corresponding multiple sample mastery evaluation results. The framework based on a feedforward neural network is trained using training data until convergence, thus obtaining the mastery evaluator.
9. The adaptive learning path planning method for knowledge networks as described in claim 1, characterized in that, The process further includes: assessing the mastery of the initial learning path and adaptively optimizing the initial learning path based on the mastery assessment results to determine the target learning path; and also includes: Get the preset feedback optimization window; The target learning path is analyzed for path congestion in the preset feedback optimization window, and learning content is pushed based on the analysis results.
10. An adaptive learning path planning system for knowledge networks, characterized in that, The steps for implementing the adaptive learning path planning method for knowledge networks according to any one of claims 1 to 9 include: The knowledge network construction module is used to construct a knowledge network with knowledge points as nodes and the relationships between knowledge points from beginning to end as directed edges. The subgraph extraction module is used to set a target node and extract multiple subgraphs from the knowledge network that can reach the target node; The initial learning path acquisition module is used to perform a breadth-first traversal of the multiple subgraphs based on path length and path congestion level to obtain the initial learning path; The target learning path determination module is used to assess the mastery of the initial learning path and adaptively optimize the initial learning path based on the mastery assessment results to determine the target learning path.