Learning ability evaluation model construction method for personalized learning of online learning platform

Abstract

The invention belongs to the technical field of data processing, and provides a learning ability evaluation model construction method for personalized learning of an online learning platform, which comprises the following steps: firstly, collecting behavior data and learning resource metadata of a user in a learning process; an individual behavior heterogeneous graph, a knowledge system hypergraph and a knowledge point sequential graph are constructed; then, a learning scene attention matrix is preset, and a context sensing mechanism is determined in combination with the knowledge point sequential diagram; and finally, according to a context sensing mechanism, the individual behavior heterogeneous graph and the knowledge system hypergraph, constructing a learning ability evaluation model. According to the method, by constructing the individual behavior heterogeneous graph, the knowledge system hypergraph and the knowledge point sequential graph, behavior data generated by the user in the learning process is deeply fused with metadata such as the type and difficulty of learning resources and associated knowledge points, accurate modeling of a complex interaction relation between the user-resource-knowledge points is achieved, and the user-resource-knowledge point interaction relation is established. Systematicness and dynamics of a knowledge system are comprehensively shown, and the accuracy of learning ability evaluation is effectively improved.

Description

Technical Field

[0001] The present invention belongs to the field of data processing technology, and in particular relates to a method for constructing a learning ability assessment model for personalized learning on an online learning platform. Background Art

[0002] As the wave of digital education sweeps the globe, online learning platforms have become a vital channel for millions of learners to acquire knowledge. Users exhibit significant differences in learning progress, knowledge base, and cognitive styles, leading to an increasing demand for personalized learning. To address this need, learning ability assessment models have emerged. These models aim to analyze learner behavioral data to accurately assess their knowledge mastery and learning ability, thereby providing tailored learning solutions. However, current technological development still faces numerous bottlenecks. Existing learning ability assessment models mostly rely on single-modal data, the rich information generated during the learning process. They employ fixed weight distribution in feature processing, making them inflexible and unable to adapt to the personalized learning styles of different learners. Furthermore, knowledge relationship modeling is often limited to simple bipartite graph structures, which inadequately capture the complexity of knowledge systems and lack awareness of high-level dependencies between learning scenarios and knowledge points. These issues result in insufficient data value, making models incapable of adapting to diverse learning scenarios, and providing superficial representations of knowledge systems, failing to accurately grasp the intricate connections between learner behavior and knowledge structures. Summary of the Invention

[0003] In view of this, the present invention provides a method for constructing a learning ability assessment model for personalized learning on an online learning platform, aiming to solve the problem in the prior art that the existing learning ability assessment model cannot accurately calculate the complex relationship between learner behavior and knowledge structure.

[0004] A first aspect of an embodiment of the present invention provides a method for constructing a learning ability assessment model for personalized learning on an online learning platform, comprising:

[0005] Collect user behavior data and learning resource metadata during the learning process;

[0006] Based on behavioral data and learning resource metadata, construct individual behavior heterogeneous graphs, knowledge system hypergraphs, and knowledge point sequential graphs;

[0007] Preset the learning scenario attention matrix and determine the context perception mechanism based on the learning scenario attention matrix and the knowledge point sequential diagram;

[0008] A learning ability assessment model is constructed based on the context-aware mechanism, individual behavior heterogeneous graph and knowledge system hypergraph.

[0009] In one possible implementation, based on behavioral data and learning resource metadata, a heterogeneous graph of individual behaviors, a hypergraph of knowledge systems, and a sequential graph of knowledge points are constructed, including:

[0010] Determine the logical relationships between heterogeneous graph nodes, hypergraph nodes, and knowledge points based on learning resource metadata;

[0011] Based on the behavioral data, determine the edges of the heterogeneous graph and construct the individual behavior heterogeneous graph by combining the nodes of the heterogeneous graph;

[0012] Determine the edges of the hypergraph based on the logical relationship between knowledge points, and build a knowledge system hypergraph by combining hypergraph nodes;

[0013] According to the knowledge system hypergraph and the logical relationship between knowledge points, a knowledge point sequential graph is established.

[0014] In one possible implementation, based on the behavioral data, edges of the heterogeneous graph are determined, and the individual behavior heterogeneous graph is constructed by combining the nodes of the heterogeneous graph, including:

[0015] Define user-learning resource edges, user-knowledge point edges, and learning resource-knowledge point edges;

[0016] Set a first-order time decay function on the edge between user and learning resource, and set an exponential decay function on the edge between user and knowledge point;

[0017] Connect the user-learning resource edges, user-knowledge point edges, and learning resource-knowledge point edges to each heterogeneous graph node to obtain an individual behavior heterogeneous graph.

[0018] In one possible implementation, the edges of the hypergraph are determined based on the logical relationships between knowledge points, and the knowledge system hypergraph is constructed by combining the hypergraph nodes, including:

[0019] According to the logical relationship of knowledge points, determine the hierarchical inclusion relationship, horizontal parallel relationship and vertical progressive relationship of knowledge points;

[0020] According to the hierarchical inclusion relationship, horizontal parallel relationship and vertical progressive relationship, the knowledge point matching score is calculated as the edge attribute value of the hypergraph;

[0021] Connect the edges of the hypergraph to each hypergraph node to obtain the knowledge system hypergraph.

[0022] In one possible implementation, a knowledge point sequential graph is established based on the knowledge system hypergraph and the logical relationship between knowledge points, including:

[0023] Extracting directed hyperedges with vertical progressive relationships from the knowledge system hypergraph, and screening the extracted directed hyperedges based on the logical relationships of knowledge points;

[0024] The filtered directed hyperedges are converted into directed edges of the knowledge point sequential graph, where the direction of the directed edge is consistent with the direction of the hyperedge, and the edge weight retains the dependency strength, thereby obtaining the knowledge point sequential graph.

[0025] In one possible implementation, a context-aware mechanism is determined based on the learning scenario attention matrix and the knowledge point sequential graph, including:

[0026] According to the learning scenario attention matrix, determine the attention weight of each knowledge point in the knowledge point sequential graph in different scenarios;

[0027] The attention weight of each knowledge point in different scenarios is added as a node attribute to the knowledge point sequential graph to form a context-aware mechanism.

[0028] In one possible implementation, a learning scenario attention matrix is ​​preset, including:

[0029] Setting multiple three-dimensional feature vectors; wherein the three-dimensional feature vectors include user state features, resource attribute features, and time environment features; each three-dimensional feature vector is used for a type of learning scenario;

[0030] Based on multiple three-dimensional feature vectors, a learning scene attention matrix is ​​formed.

[0031] In one possible implementation, a learning ability assessment model is constructed based on the context-aware mechanism, the individual behavior heterogeneous graph, and the knowledge system hypergraph, including:

[0032] Taking the context-aware mechanism as the scene sub-network, the individual behavior heterogeneous graph as the behavior sub-network, and the knowledge system hypergraph as the knowledge sub-network, a cross-view graph neural network is constructed as a learning ability evaluation model.

[0033] In one possible implementation, the context-aware mechanism is used as a scene sub-network, the individual behavior heterogeneous graph is used as a behavior sub-network, and the knowledge system hypergraph is used as a knowledge sub-network to construct a cross-view graph neural network as a learning ability evaluation model, including:

[0034] The context-aware mechanism is used as the scene sub-network, the individual behavior heterogeneous graph is used as the behavior sub-network, and the knowledge system hypergraph is used as the knowledge sub-network. At the same time, the contrast loss function based on positive and negative sample pairs is used as the total network loss. The three-view embedding is weightedly fused through the self-attention mechanism to obtain a learning ability evaluation model.

[0035] In one possible implementation, the user's behavior data and learning resource metadata during the learning process are collected, including:

[0036] Collect raw data from online learning platforms;

[0037] The raw data is preprocessed to obtain behavioral data and learning resource metadata.

[0038] The method for constructing a learning ability assessment model for personalized learning on an online learning platform provided by an embodiment of the present invention first collects the behavioral data and learning resource metadata of the user during the learning process; then, based on the behavioral data and learning resource metadata, constructs an individual behavioral heterogeneous graph, a knowledge system hypergraph, and a knowledge point sequential graph; then, presets a learning scenario attention matrix, and determines a context perception mechanism based on the learning scenario attention matrix and the knowledge point sequential graph; finally, constructs a learning ability assessment model based on the context perception mechanism, the individual behavioral heterogeneous graph, and the knowledge system hypergraph. By constructing an individual behavioral heterogeneous graph, a knowledge system hypergraph, and a knowledge point sequential graph, the present invention deeply integrates the behavioral data generated by the user during the learning process with metadata such as the type, difficulty, and associated knowledge points of the learning resource, thereby achieving accurate modeling of the complex interactive relationship between "user-resource-knowledge point", comprehensively displaying the systematicity and dynamics of the knowledge system, and effectively improving the accuracy of learning ability assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 This is an application scenario diagram of the method for constructing a learning ability assessment model for personalized learning on an online learning platform provided by an embodiment of the present invention;

[0041] Figure 2 This is a flowchart of a method for constructing a learning ability assessment model for personalized learning on an online learning platform provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In the following description, specific details such as particular system structures and techniques are provided for purposes of illustration, not limitation, to facilitate a thorough understanding of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.

[0043] Figure 1 This is an application scenario diagram of the method for constructing a learning ability assessment model for personalized learning on an online learning platform provided by an embodiment of the present invention. Figure 1 As shown, the scenario includes a server 11 and at least one terminal 12 .

[0044] Server 11 is equipped with an online learning platform. Students access server 11 through terminal 12 to study. Server 11 records students' learning content and behavior in real time. Based on the recorded content and the server's knowledge base, it establishes a personal learning ability assessment model for students, thereby better dynamically evaluating their learning ability. This provides more accurate assessment content for each student, allowing them to develop better learning plans or recommend more suitable courses.

[0045] Figure 2 This is a flow chart of the implementation of the method for constructing a learning ability assessment model for personalized learning on an online learning platform provided by an embodiment of the present invention. Figure 2 As shown, in some embodiments, a method for constructing a learning ability assessment model for personalized learning on an online learning platform includes:

[0046] S210, collecting user behavior data and learning resource metadata during the learning process;

[0047] In the embodiment of the present invention, behavioral data refers to the dynamic interaction data generated by users during the learning process, specifically including learning progress data, learning time, interaction records, operation behaviors, evaluation feedback, etc.

[0048] Learning resource metadata refers to static data that describes the inherent properties and structure of learning resources, including:

[0049] Basic attributes: resource type (such as courses, e-books, audio, journals, AR / VR content), format (such as Epub files, panoramic videos), difficulty level, and applicable knowledge areas.

[0050] Knowledge point association: the knowledge points contained in the resource and its logical relationship with other resources (such as hierarchical inclusion, horizontal parallelism, and vertical progressive relationship, see the description of knowledge system hypergraph construction).

[0051] Metadata standards: resource ID, title, author, publication time, classification tags (such as system-preset course templates and special resource tags).

[0052] Structured information: the three-level directory structure of the course outline, the directory hierarchy of the e-book, and the triggering conditions of AR resources (such as 2D / 3D object scanning and recognition rules).

[0053] Evaluation-related attributes: the knowledge point matching score corresponding to the resource, and the advancement conditions in the knowledge advancement map (such as stage test papers and capability analysis reports).

[0054] In some embodiments, collecting the user's behavior data and learning resource metadata during the learning process includes: collecting raw data from the online learning platform; and preprocessing the raw data to obtain the behavior data and learning resource metadata.

[0055] In this embodiment of the present invention, the collected data is first processed for missing values, outlier filtering, and data deduplication. Key features are then extracted from the raw data and converted into structured data that can be used for model building. Specifically, different methods can be used for structuring data of different modalities and sources, and this is not limited here.

[0056] S220, constructing individual behavior heterogeneous graph, knowledge system hypergraph and knowledge point sequential graph based on behavior data and learning resource metadata;

[0057] S230, presetting a learning scenario attention matrix, and determining a context perception mechanism based on the learning scenario attention matrix and the knowledge point sequential graph;

[0058] S240, constructs a learning ability evaluation model based on the context-aware mechanism, individual behavior heterogeneous graph and knowledge system hypergraph.

[0059] In an embodiment of the present invention, by constructing an individual behavior heterogeneous graph, a knowledge system hypergraph and a knowledge point sequential graph, the behavioral data generated by the user during the learning process is deeply integrated with metadata such as the type, difficulty, and related knowledge points of the learning resources, thereby achieving accurate modeling of the complex interactive relationship between "user-resource-knowledge point", comprehensively displaying the systematic and dynamic nature of the knowledge system, and effectively improving the accuracy of learning ability assessment.

[0060] In some embodiments, an individual behavior heterogeneous graph, a knowledge system hypergraph and a knowledge point sequential graph are constructed based on behavior data and learning resource metadata, including: determining the heterogeneous graph nodes, hypergraph nodes and logical relationships of knowledge points based on learning resource metadata; determining the edges of the heterogeneous graph based on behavior data, and constructing an individual behavior heterogeneous graph in combination with the heterogeneous graph nodes; determining the edges of the hypergraph based on the logical relationships of knowledge points, and constructing a knowledge system hypergraph in combination with the hypergraph nodes; and establishing a knowledge point sequential graph based on the knowledge system hypergraph and the logical relationships of knowledge points.

[0061] In an embodiment of the present invention, when determining heterogeneous graph nodes based on learning resource metadata, three types of nodes are clearly included: user nodes correspond to various types of platform users (such as students, teachers, and administrators), and their attributes can be associated with user tags, learning ability assessment results, etc.; learning resource nodes cover multiple types such as courses, e-books, audio, journals, AR / VR resources, etc., and need to be annotated with metadata such as resource ID, type, difficulty, and label; knowledge point nodes are the smallest knowledge units extracted from resource metadata, such as the third-level directory knowledge points of the course outline, and the knowledge point ID, the field to which it belongs, and the associated resources need to be recorded.

[0062] Hypergraph nodes are constructed with knowledge points as the core, supplemented by resource nodes. Knowledge point nodes need to be further annotated with hierarchical attributes (such as elementary / intermediate / advanced) and logical relationship types (hierarchical inclusion / horizontal parallel / vertical progression). For example, in the "knowledge advancement map subsystem", resource packages in different knowledge fields need to be organized according to the knowledge point hierarchy. Hypergraph nodes can be directly mapped to stage knowledge points in the knowledge advancement map, and their attributes need to match the advancement conditions (such as "completing the previous knowledge points before unlocking the subsequent content"). Resource nodes, as the carriers of knowledge points, are associated with corresponding knowledge point nodes to strengthen the mapping relationship between the knowledge system and resources.

[0063] Logical relationships between knowledge points are extracted from structured information in learning resource metadata and include three core types: Hierarchical inclusion relationships reflect the subordinate structure of knowledge points, such as "Mathematics" including "Algebra" and "Geometry." Horizontal parallel relationships refer to the parallel association of knowledge points at the same level, such as "Python" and "Java" under "Programming Languages." Vertical progressive relationships indicate the forward and backward dependencies of knowledge points, such as "Basics of Calculus" → "Multivariate Calculus." These relationships are modeled based on the course syllabus's directory structure, the progression rules of the knowledge progression map, and the knowledge point association annotations in the resource metadata.

[0064] In some embodiments, based on behavioral data, the edges of the heterogeneous graph are determined, and an individual behavioral heterogeneous graph is constructed in combination with the heterogeneous graph nodes, including: defining user-learning resource edges, user-knowledge point edges, and learning resource-knowledge point edges; setting a first-order time decay function on the user-learning resource edges, and setting an exponential decay function on the user-knowledge point edges; connecting the user-learning resource edges, user-knowledge point edges, and learning resource-knowledge point edges to each heterogeneous graph node to obtain an individual behavioral heterogeneous graph.

[0065] In this embodiment of the present invention, the user-learning resource edge represents a user's direct actions on a learning resource, such as accessing, downloading, adding to favorites, caching resources, etc. Specifically, this can be obtained by querying user operation records in the behavior log, such as "user ID + resource ID + operation time + operation type (browsing / downloading, etc.)".

[0066] In this embodiment of the present invention, the user-knowledge point edge reflects the user's interaction with the knowledge point, such as the learning time of the knowledge point, the correct answer rate, the number of wrong questions saved, etc. It is specifically determined by the evaluation data during the learning process, such as "user ID + knowledge point ID + mastery score + answer time".

[0067] In this embodiment of the present invention, the learning resource-knowledge point edge describes the static association between learning resources and knowledge points, such as the list of knowledge points included in a course or the knowledge points corresponding to an e-book chapter. Specifically, it is determined by the structured information in the resource metadata, such as the knowledge points corresponding to the third-level directory of the course syllabus.

[0068] In an embodiment of the present invention, the first-order time decay function is:

[0069]

[0070] Among them, w u-r is the edge weight, t is the operation time, T is the current time, and λ is the decay coefficient. The weight decays linearly with time, emphasizing that the user's recent operation on the resource can better reflect the current learning status. For example, the weight of the resource downloaded by the user 1 day ago is 0.9, and it decays to 0.5 after 7 days.

[0071] In an embodiment of the present invention, the exponential decay function is:

[0072] w u-k (t) = e -λ(T-t)

[0073] Among them, w u-k The weight decays exponentially over time, reinforcing the impact of short-term memory on knowledge mastery. For example, the weight of a knowledge point mastered by a user one day ago is 0.9, and after seven days it decays to 0.5 (λ = 0.3), reflecting the forgetting law faster than first-order decay.

[0074] Whenever a user generates a new behavior, the decay weight of the corresponding edge is recalculated to ensure that the graph structure reflects the user's latest learning dynamics in real time, thereby ensuring the continuous update of the cross-view graph neural network.

[0075] In some embodiments, the edges of the hypergraph are determined based on the logical relationships of the knowledge points, and a knowledge system hypergraph is constructed in combination with the hypergraph nodes, including: determining the hierarchical inclusion relationship, horizontal parallel relationship, and vertical progressive relationship of the knowledge points based on the logical relationships of the knowledge points; calculating the knowledge point matching score as the edge attribute value of the hypergraph based on the hierarchical inclusion relationship, horizontal parallel relationship, and vertical progressive relationship; and connecting the edges of the hypergraph to each hypergraph node to obtain a knowledge system hypergraph.

[0076] In the embodiment of the present invention, three core types can be determined based on the logical relationship of knowledge points:

[0077] Hierarchical inclusion relationship: reflects the subordinate structure of knowledge points, such as "mathematics" includes "algebra" and "geometry".

[0078] Horizontal parallel relationship: refers to the parallel association of knowledge points at the same level, such as "Python" and "Java" under "Programming Language". Its modeling basis includes the same-level labels in the resource classification system and the parallel design of course modules.

[0079] Vertical progressive relationship: indicates the order of dependency of knowledge points, such as "Basics of Calculus" → "Multivariate Calculus", which corresponds to the advanced condition of "completing the previous knowledge points to unlock the subsequent content" in the knowledge advancement map and the mentioned "vertical progressive directed hyperedge".

[0080] Based on three types of logical relationships, a knowledge point matching score is calculated as a hypergraph edge attribute value. This quantifies the closeness of the logical connection between knowledge points (with a value between 0 and 1). For example, the hierarchical inclusion score for "Algebra" and "Mathematics" can be set to 0.9, the horizontal parallel score for "Python" and "Java" to 0.7, and the vertical progressive score for "Basic Calculus" and "Multivariate Calculus" to 0.95. This score is determined by methods such as the depth of the course syllabus hierarchy, the frequency of co-occurrence of user learning behaviors, or expert annotation to ensure that it reflects the true strength of the logical connection.

[0081] When constructing a knowledge system hypergraph, knowledge points are used as core nodes, supplemented by resource nodes as knowledge carriers:

[0082] Node attributes: Knowledge point nodes are marked with ID, field, level and logical relationship type, and resource nodes are associated with corresponding knowledge points.

[0083] Edge connection rules:

[0084] The hierarchy includes: parent knowledge points are connected to child knowledge points through directed hyperedges (such as "mathematics" → {"algebra" "geometry"}), and the edge weight is the matching score.

[0085] Horizontal parallel: Knowledge points at the same level are connected by undirected hyperedges or bidirectional directed edges (such as ), and the edge weight is the matching score.

[0086] Vertical progression: The preceding knowledge point points to the subsequent knowledge point through a directed hyperedge (such as "Basics of Calculus" → "Multivariate Calculus"), and the edge weight is the dependency strength score.

[0087] By modeling "one-to-many" and "many-to-many" relationships through high-order edges, the systematic and dynamic nature of the knowledge system is fully demonstrated, providing a global modeling foundation at the knowledge level for learning ability assessment.

[0088] In some embodiments, a knowledge point sequential graph is established based on the knowledge system hypergraph and the logical relationship between knowledge points, including: extracting directed hyperedges with vertical progressive relationships from the knowledge system hypergraph, and filtering the extracted directed hyperedges according to the logical relationship between knowledge points; converting the filtered directed hyperedges into directed edges of the knowledge point sequential graph, wherein the edge direction of the directed edge is consistent with the hyperedge direction, and the edge weight retains the dependency strength, thereby obtaining a knowledge point sequential graph.

[0089] In the embodiment of the present invention, when extracting directed hyperedges of vertical progressive relationships from the knowledge system hypergraph, it is necessary to focus on the sequential relationship of the front-to-back dependencies between knowledge points. This type of directed hyperedge represents the logic of "previous knowledge points must be mastered first, and subsequent knowledge points can be learned" (such as "Basics of Calculus → Multivariate Calculus"), and it is necessary to exclude hierarchical inclusion (such as "Mathematics → Algebra") and horizontal parallelism (such as ) relationship. The extraction basis is the directed hyperedges marked with the "vertical progressive" type in the knowledge system hypergraph, and their matching scores reflect the strength of knowledge point dependency.

[0090] There are two conditions for filtering directed hyperedges: first, logical relationship conformity, retaining only edges clearly marked as "vertical progression"; and second, a dependency strength threshold, setting a matching score ≥ 0.7 (or other reasonable threshold) to filter out weakly associated edges. For example, in the hypergraph, "Basics of Calculus → Multivariate Calculus" (score 0.95) meets the conditions and can be retained, while "Algebra → Geometry" (horizontally parallel, score 0.6) is filtered out. Through screening, it is ensured that the knowledge point sequential graph contains only strong dependencies that can guide the learning order, providing accurate data support for learning path planning.

[0091] When converting the selected directed hyperedges into directed edges of the knowledge point sequential graph, the principles of consistent direction and weight preservation must be followed: the hyperedge direction is directly mapped to the direction of the directed edge of the sequential graph (such as "knowledge point A→knowledge point B"), and its matching score is used as the edge weight to represent the dependency strength. The constructed knowledge point sequential graph is a directed acyclic graph (DAG), which supports the generation of linear learning paths (such as "knowledge point A→knowledge point B→knowledge point C") through topological sorting, and can be applied to two major scenarios: one is learning plan navigation, which automatically generates personalized learning paths with stage assessments based on the sequential graph; the other is the knowledge advancement map, which uses the sequence of knowledge points in the sequential graph as an advancement condition, such as "subsequent content can only be unlocked after completing the prerequisite knowledge points and achieving the required grades."

[0092] In some embodiments, a context-aware mechanism is determined based on the learning scenario attention matrix and the knowledge point sequential graph, including: determining the attention weight of each knowledge point in the knowledge point sequential graph in different scenarios based on the learning scenario attention matrix; adding the attention weight of each knowledge point in different scenarios as a node attribute to the knowledge point sequential graph to form a context-aware mechanism.

[0093] In an embodiment of the present invention, the construction of the learning scenario attention matrix is ​​based on three-dimensional feature vectors, and each vector corresponds to a type of learning scenario. Among them, the three-dimensional feature vector includes user status characteristics (such as learning progress, points, learning ability assessment level), resource attribute characteristics (such as resource type, difficulty, knowledge point association) and time environment characteristics (such as learning period, duration). By presetting a plurality of such three-dimensional feature vectors, a learning scenario attention matrix can be formed to quantify the importance differences of knowledge points in different scenarios. For example, for the "novice user + primary resources + night learning" scenario, the matrix can give basic knowledge points a higher attention weight, reflecting that the user's need for basic knowledge in this scenario is more urgent.

[0094] Based on the learning scenario attention matrix, it is necessary to calculate the attention weight of each knowledge point in the knowledge point sequential graph in different scenarios. Specifically, by mapping the matrix with the node features of the knowledge point sequential graph, the weight value of each knowledge point in the "user status-resource attribute-time environment" combination scenario can be determined. For example, the weight of the knowledge point "variable definition" in the "novice user + Python course + weekday daytime" scenario is 0.8, while the weight in the "advanced user + Java framework + weekend night" scenario is only 0.3. This type of weight calculation process ensures that the model can perceive the impact of the user's current scenario on the importance of the knowledge point.

[0095] After the scene attention weight of each knowledge point is added as a node attribute to the knowledge point sequential graph, a context-aware mechanism is formed. At this point, each node in the knowledge point sequential graph not only contains the logical relationship and dependency strength of the knowledge point itself, but also carries scene-sensitive attention information. For example, in user learning path planning, the system can automatically increase the weight of the corresponding knowledge point according to the current scenario, and give priority to recommending related resources or evaluation content. This mechanism dynamically adjusts the priority of knowledge points by integrating scenarios, achieving the goal of enhancing the model's adaptability to diverse learning scenarios, and providing key input for the subsequent modeling of cross-view graph neural networks (scenario subnetworks).

[0096] In some embodiments, a learning scenario attention matrix is ​​preset, including: setting multiple three-dimensional feature vectors; wherein the three-dimensional feature vectors include user state features, resource attribute features and time environment features; each three-dimensional feature vector is used for a type of learning scenario; and based on multiple three-dimensional feature vectors, a learning scenario attention matrix is ​​formed.

[0097] In this embodiment of the present invention, the three-dimensional feature vector is constructed based on the core influencing factors of the learning scenario and includes three key features:

[0098] User status features: covers the user's learning progress on the platform (such as course completion rate, accumulated points in the points system, learning ability assessment results (such as level, scores in each dimension) and user tags (such as novice / advanced user).

[0099] Resource attribute characteristics: including the type of learning resources (such as courses, e-books, AR content), difficulty level (beginner / intermediate / advanced), knowledge point association (such as the knowledge field, prerequisite knowledge points) and label classification (such as "high-frequency test points" and "special resources").

[0100] Temporal environment characteristics: involving the time period when learning behavior occurs (such as weekday daytime / weekend nighttime), duration (such as the duration of a single study session) and periodic patterns (such as the pre-exam sprint stage and daily study).

[0101] The mapping between three-dimensional feature vectors and learning scenarios is achieved by combining three types of features, where each vector uniquely identifies a type of learning scenario:

[0102] Scenario 1: Novice users + beginner courses + fragmented time

[0103] Vector example: [User Level = Novice, Resource Difficulty = Elementary, Time = Weekday Daytime (≤30 minutes)] is suitable for scenarios where users use fragmented time to learn basic knowledge points.

[0104] Scenario 2: Advanced users + complex courses + a whole block of time

[0105] Vector example: [User Level = Advanced, Resource Type = Professional Course, Time = Weekend Afternoon (≥ 2 hours)] corresponds to scenarios that require in-depth learning of complex knowledge points.

[0106] Scenario 3: Exam Prep + High-Frequency Exam Resources + Pre-Exam Stage

[0107] Vector example: [Learning Objective = Exam Preparation, Resource Tag = High-Frequency Exam Points, Time = 1 Week Before Exam], which matches the scenario requirements of a intensive review before the exam.

[0108] The learning scenario attention matrix is ​​constructed using multiple three-dimensional eigenvectors as row vectors to form a multidimensional matrix. Each row corresponds to a learning scenario, and the columns are eigenvalues ​​or weight parameters for user status, resource attributes, and temporal context. This matrix, through pre-set feature combinations, provides a basis for scenario classification for subsequent context-aware mechanisms. This allows the model to extract corresponding eigenvectors from the matrix based on the user's current scenario and dynamically adjust the attention weights of each node in the knowledge point sequential graph, achieving the goal of "differentiated assessment of the importance of knowledge points in different scenarios."

[0109] In some embodiments, a learning ability evaluation model is constructed based on the context-aware mechanism, the individual behavior heterogeneous graph and the knowledge system hypergraph, including: using the context-aware mechanism as the scene sub-network, the individual behavior heterogeneous graph as the behavior sub-network, and the knowledge system hypergraph as the knowledge sub-network to construct a cross-view graph neural network as a learning ability evaluation model.

[0110] In some embodiments, a context-aware mechanism is used as a scene sub-network, an individual behavior heterogeneous graph is used as a behavior sub-network, and a knowledge system hypergraph is used as a knowledge sub-network to construct a cross-view graph neural network as a learning ability evaluation model, including: using the context-aware mechanism as a scene sub-network, the individual behavior heterogeneous graph as a behavior sub-network, and the knowledge system hypergraph as a knowledge sub-network, and at the same time using the contrast loss function based on positive and negative sample pairs as the total network loss, and performing weighted fusion of the three-view embeddings through a self-attention mechanism to obtain a learning ability evaluation model.

[0111] In this embodiment of the present invention, the construction of the learning ability assessment model is based on three sub-networks, corresponding to the modeling of learning scenarios, user behaviors, and knowledge systems respectively:

[0112] The context-aware mechanism serves as a scenario subnetwork: This subnetwork assigns context-sensitive attention weights to nodes in the sequential graph of knowledge points by learning a scenario-based attention matrix. For example, it generates three-dimensional feature vectors based on user status, resource attributes, and temporal context, dynamically adjusting the importance of knowledge points in different scenarios. Its core function is to enable the model to perceive changes in learning scenarios and adapt to diverse learning needs.

[0113] The individual behavior heterogeneous graph serves as a behavioral subnetwork: This subnetwork models dynamic user behavior data using three types of edges (user, resource, and knowledge point) and a time decay function, depicting the real-time interaction between users, resources, and knowledge points. For example, recently downloaded resources and frequently accessed knowledge points are given higher weights, reflecting a user's current learning focus.

[0114] The knowledge system hypergraph serves as a knowledge subnetwork: This subnetwork constructs a hypergraph structure based on the hierarchical inclusion, horizontal parallelism, and vertical progression of knowledge points. The strength of the logical associations between knowledge points is quantified through matching scores. For example, the vertical progression edge weight between "Basics of Calculus" and "Multivariate Calculus" is 0.95, reflecting a strong dependency order and supporting systematic modeling of the knowledge system.

[0115] The integration mechanism of cross-view graph neural networks is implemented through the following steps:

[0116] Parallel input of multiple sub-networks: The scene sub-network (context-aware mechanism), behavior sub-network (individual behavior heterogeneous graph), and knowledge sub-network (knowledge system hypergraph) are respectively input as independent views into different branches of the graph neural network to extract the feature representation of each view.

[0117] Self-Attention Fusion: The self-attention mechanism is used to weight the embedding vectors of the three views and automatically learn the importance of different views in the evaluation. For example, when evaluating the learning ability of novice users, the recent operation records of the behavior sub-network and the "novice scene" features of the scene sub-network may be given higher weights.

[0118] Contrastive loss function optimization: A contrastive loss function based on positive and negative sample pairs is introduced as the total network loss. This forces the feature representations of similar samples to be closer in the embedding space and heterogeneous samples to be further apart, thereby improving the model's ability to distinguish differences in user learning abilities.

[0119] This invention integrates heterogeneous data (behavior + knowledge + scenario) through graph neural networks, breaking through the limitations of traditional models that rely on single-modal data, achieving accurate modeling of the complex relationship between "users-resources-knowledge points", and effectively improving the accuracy of learning ability assessment.

[0120] It should be understood that the size of the serial numbers of the steps in the above embodiments 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 invention.

[0121] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A method for constructing a learning ability assessment model for personalized learning on an online learning platform, characterized in that: include: Collect user behavior data and learning resource metadata during the learning process; Constructing an individual behavior heterogeneous graph, a knowledge system hypergraph, and a knowledge point sequential graph based on the behavior data and the learning resource metadata; Presetting a learning scenario attention matrix, and determining a context perception mechanism based on the learning scenario attention matrix and a knowledge point sequential graph; A learning ability evaluation model is constructed based on the context-aware mechanism, the individual behavior heterogeneous graph and the knowledge system hypergraph.

2. The method for constructing a learning ability assessment model for personalized learning on an online learning platform according to claim 1, characterized in that: Based on the behavior data and the learning resource metadata, an individual behavior heterogeneous graph, a knowledge system hypergraph, and a knowledge point sequential graph are constructed, including: Determining the logical relationships among heterogeneous graph nodes, hypergraph nodes, and knowledge points based on the learning resource metadata; Determine edges of a heterogeneous graph based on the behavior data, and construct an individual behavior heterogeneous graph by combining the nodes of the heterogeneous graph; Determine the edges of the hypergraph according to the logical relationship of the knowledge points, and construct a knowledge system hypergraph by combining the hypergraph nodes; According to the knowledge system hypergraph and the logical relationship between the knowledge points, a knowledge point sequential graph is established.

3. The method for constructing a learning ability assessment model for personalized learning on an online learning platform according to claim 2, characterized in that: Determining edges of a heterogeneous graph based on the behavior data, and constructing an individual behavior heterogeneous graph in combination with the nodes of the heterogeneous graph, including: Define user-learning resource edges, user-knowledge point edges, and learning resource-knowledge point edges; Setting a first-order time decay function on the edge between the user and the learning resource, and setting an exponential decay function on the edge between the user and the knowledge point; Connect the user-learning resource edges, user-knowledge point edges, and learning resource-knowledge point edges to each heterogeneous graph node to obtain an individual behavior heterogeneous graph.

4. The method for constructing a learning ability assessment model for personalized learning on an online learning platform according to claim 2, characterized in that: Determine the edges of the hypergraph based on the logical relationship of the knowledge points, and construct a knowledge system hypergraph by combining the hypergraph nodes, including: According to the logical relationship of knowledge points, determine the hierarchical inclusion relationship, horizontal parallel relationship and vertical progressive relationship of knowledge points; According to the hierarchical inclusion relationship, horizontal parallel relationship and vertical progressive relationship, the knowledge point matching score is calculated as the edge attribute value of the hypergraph; Connect the edges of the hypergraph to each hypergraph node to obtain the knowledge system hypergraph.

5. The method for constructing a learning ability assessment model for personalized learning on an online learning platform according to claim 2, characterized in that: According to the knowledge system hypergraph and the logical relationship between the knowledge points, a knowledge point sequential graph is established, including: Extracting directed hyperedges with vertical progressive relationships from the knowledge system hypergraph, and screening the extracted directed hyperedges according to the logical relationships of the knowledge points; The filtered directed hyperedges are converted into directed edges of the knowledge point sequential graph, where the direction of the directed edge is consistent with the direction of the hyperedge, and the edge weight retains the dependency strength, thereby obtaining the knowledge point sequential graph.

6. The method for constructing a learning ability assessment model for personalized learning on an online learning platform according to claim 1, characterized in that: According to the learning scenario attention matrix and the knowledge point sequential diagram, a context perception mechanism is determined, including: Determining the attention weight of each knowledge point in the knowledge point sequential graph in different scenarios according to the learning scenario attention matrix; The attention weight of each knowledge point in different scenarios is added as a node attribute to the knowledge point sequential graph to form a context-aware mechanism.

7. The method for constructing a learning ability assessment model for personalized learning on an online learning platform according to claim 6, characterized in that: Preset learning scenario attention matrix, including: Setting a plurality of three-dimensional feature vectors; wherein the three-dimensional feature vectors include user state features, resource attribute features, and time environment features; each three-dimensional feature vector is used for a type of learning scenario; Based on multiple three-dimensional feature vectors, a learning scene attention matrix is ​​formed.

8. The method for constructing a learning ability assessment model for personalized learning on an online learning platform according to claim 1, characterized in that: Based on the context-aware mechanism, the individual behavior heterogeneous graph, and the knowledge system hypergraph, a learning ability assessment model is constructed, including: The context-aware mechanism is used as the scene sub-network, the individual behavior heterogeneous graph is used as the behavior sub-network, and the knowledge system hypergraph is used as the knowledge sub-network to construct a cross-view graph neural network as a learning ability evaluation model.

9. The method for constructing a learning ability assessment model for personalized learning on an online learning platform according to claim 8, characterized in that: The context-aware mechanism is used as the scene sub-network, the individual behavior heterogeneous graph is used as the behavior sub-network, and the knowledge system hypergraph is used as the knowledge sub-network to construct a cross-view graph neural network as a learning ability evaluation model, including: The context-aware mechanism is used as the scene sub-network, the individual behavior heterogeneous graph is used as the behavior sub-network, and the knowledge system hypergraph is used as the knowledge sub-network. At the same time, the contrast loss function based on positive and negative sample pairs is used as the total network loss. The three-view embedding is weightedly fused through the self-attention mechanism to obtain a learning ability evaluation model.

10. The method for constructing a learning ability assessment model for personalized learning on an online learning platform according to any one of claims 1 to 9, characterized in that: Collect user behavior data and learning resource metadata during the learning process, including: Collect raw data from online learning platforms; The raw data is preprocessed to obtain behavior data and learning resource metadata.

Images