Method and system for comprehensive evaluation of education based on intelligent education AI model

Abstract

The present application belongs to the technical field of intelligent education, and discloses a method and system for comprehensive education evaluation based on an intelligent education AI model, comprising: gathering multi-modal cognitive data and learning behavior trajectories, coupling multiple sources to obtain an education interaction information spectrum; decoding cognitive structure to obtain a learning behavior cognitive map, constructing a knowledge network topology to obtain a cognitive development topographic map; performing cognitive fault detection to obtain a cognitive transition critical point set, performing cognitive evolution tracking to obtain a learning development trajectory map; purifying learning modes to obtain a learning style spectrum, and constructing a cognitive sensitivity analysis model; intelligently combining learning navigation schemes to obtain a set of personalized learning enhancement strategies, coordinating teaching schemes to obtain a comprehensive education evaluation report; the present application realizes intelligentization of the education evaluation process, and can accurately identify cognitive characteristics and provide personalized learning guidance.

Description

Technical Field

[0001] This invention relates to the field of smart education technology, and more specifically, to a method and system for comprehensive educational evaluation based on a smart education AI model. Background Technology

[0002] In the field of smart education, student profiling is a key step in achieving personalized teaching and precise educational intervention, and it is of great value in improving teaching quality and learning outcomes. Traditional student profiling is mainly based on static labels, including fixed academic performance ratings, single-dimensional ability assessments, and periodic summative evaluations. Although these methods are widely used, they have obvious shortcomings such as information lag, one-sided evaluation, and lack of personalization, and cannot meet the needs of modern smart education for comprehensive and dynamic student assessment.

[0003] With the development of educational data analysis technology, dynamic profiling methods based on the learning process have gradually gained attention. However, existing technologies struggle to effectively capture and analyze subtle changes and developmental trajectories in students' learning processes. Traditional evaluation systems simply categorize students into different ability levels by comparing their exam scores at fixed time points with the class or grade average, but this method fails to reflect students' true abilities and developmental potential. Significant differences in students' family environments, learning habits, and cognitive development stages often lead to biases in static evaluation results, especially when the education system temporarily elevates struggling students to a passing grade through extra tutoring. This superficial improvement masks underlying learning problems, resulting in distorted profiling. In practical teaching applications, the asynchronous development of students' abilities, the complex and ever-changing learning environment, and the subjectivity of evaluation standards further interfere with the formation of accurate profiles. Existing systems lack adaptable multi-dimensional data integration mechanisms, cannot distinguish between temporary fluctuations and substantial changes in ability, and lack the ability to sensitively identify key turning points in the learning process. This results in insufficient accuracy in providing personalized learning path planning by smart education systems, affecting the scientific and effective nature of teaching decisions and ultimately limiting the full realization of students' potential.

[0004] In view of this, the present invention proposes a method and system for comprehensive educational evaluation based on a smart education AI model to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the existing technology and to achieve the above objectives, the present invention provides the following technical solution: a method for comprehensive educational evaluation based on a smart education AI model, comprising: Step S1: Collect multimodal cognitive data and learning behavior trajectories, and perform multi-source coupling on the multimodal cognitive data and learning behavior trajectories to obtain the educational interaction information spectrum; Step S2: Decode the cognitive structure of the educational interaction information spectrum to obtain a cognitive map of learning behavior, construct the knowledge network topology based on the cognitive map of learning behavior, and then obtain a cognitive development topography map; Step S3: Collect real-time teaching interaction sequences; perform cognitive fault detection on the cognitive development topography map to obtain a set of cognitive leap critical points, and track the cognitive evolution of the real-time teaching interaction sequences based on the set of cognitive leap critical points to obtain a learning development trajectory map. Step S4: Based on the learning development trajectory map, refine the learning patterns to obtain the learning style spectrum, and construct a cognitive sensitivity analysis model based on the learning style spectrum; Step S5: Intelligently combine learning navigation schemes through the cognitive sensitivity analysis model to obtain a set of personalized learning enhancement strategies; arrange teaching coordination schemes according to the set of personalized learning enhancement strategies to obtain a comprehensive education evaluation report, and provide learning guidance based on the comprehensive education evaluation report.

[0006] Systems for comprehensive educational evaluation based on intelligent education AI models include: Data fusion module: Collects multimodal cognitive data and learning behavior trajectories, and performs multi-source coupling of multimodal cognitive data and learning behavior trajectories through contextual intelligence integration to obtain the educational interaction information spectrum; Cognitive Topology Construction Module: Decodes the cognitive structure of the educational interaction information spectrum to obtain a cognitive graph of learning behavior, constructs the knowledge network topology based on the cognitive graph of learning behavior, and then obtains a cognitive development topography map; Cognitive Tracking Module: Collects real-time teaching interaction sequences; performs cognitive fault detection on the cognitive development topography map to obtain a set of cognitive leap critical points, and tracks the cognitive evolution of real-time teaching interaction sequences based on the set of cognitive leap critical points to obtain a learning development trajectory map; Model building module: Based on the learning development trajectory map, learning patterns are purified to obtain the learning style spectrum, and a cognitive sensitivity analysis model is built based on the learning style spectrum; Evaluation and guidance module: Intelligent combination of learning navigation schemes through cognitive sensitivity analysis model to obtain a set of personalized learning enhancement strategies; teaching coordination schemes are arranged according to the set of personalized learning enhancement strategies to obtain a comprehensive education evaluation report, and learning guidance is provided based on the comprehensive education evaluation report.

[0007] The technical effects and advantages of the method and system for comprehensive educational evaluation based on a smart education AI model, as described in this invention: This invention achieves comprehensive data integration of learners' cognitive processes through the collection and multi-source coupling of multimodal cognitive data and learning behavior trajectories. It transforms scattered learning behavior data into a systematic educational interaction information spectrum, thereby enhancing the comprehensiveness of educational evaluation. Through cognitive structure decoding and knowledge network topology construction, the system can accurately map learners' cognitive development, providing a three-dimensional presentation of their cognitive structures and overcoming the limitations of the one-sidedness and subjectivity of traditional educational evaluation methods. By detecting cognitive fault lines and identifying sets of cognitive transition critical points, it can keenly capture key nodes in learners' cognitive development, track the cognitive evolution of real-time teaching interaction sequences, and generate learning development trajectory maps. This dynamic tracking mechanism significantly improves the timeliness and accuracy of educational evaluation, avoiding the static and lagging problems of traditional evaluation methods. Through learning pattern purification and learning style spectrum analysis, the system can comprehensively perceive learners' cognitive characteristics and learning preferences, constructing personalized cognitive sensitivity analysis models to effectively address the diverse needs of different learners and improve the adaptability and inclusiveness of the evaluation system. By introducing a cognitive sensitivity analysis model to intelligently combine learning navigation schemes, the system can analyze learners' cognitive states in real time, predict potential learning obstacles, and dynamically adjust teaching resources and strategies through the generation of personalized learning enhancement strategy sets, ensuring accurate learning guidance even when learning scenarios change. Based on the personalized learning enhancement strategy sets, the system coordinates teaching schemes and automatically generates comprehensive evaluation reports based on real-time learning status, guiding teaching practice and improving the educational evaluation system's ability to cope with complex learning situations, significantly enhancing learning efficiency and teaching quality. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the method for comprehensive educational evaluation based on a smart education AI model according to the present invention; Figure 2 This is a schematic diagram of the system for comprehensive educational evaluation based on the intelligent education AI model of the present invention. Detailed Implementation

[0009] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1

[0010] Please see Figure 1 As shown in this embodiment, the method for comprehensive educational evaluation based on a smart education AI model includes: Step S1: Collect multimodal cognitive data and learning behavior trajectories, and perform multi-source coupling on the multimodal cognitive data and learning behavior trajectories to obtain the educational interaction information spectrum; Step S2: Decode the cognitive structure of the educational interaction information spectrum to obtain a cognitive map of learning behavior, construct the knowledge network topology based on the cognitive map of learning behavior, and then obtain a cognitive development topography map; Step S3: Collect real-time teaching interaction sequences; perform cognitive fault detection on the cognitive development topography map to obtain a set of cognitive leap critical points, and track the cognitive evolution of the real-time teaching interaction sequences based on the set of cognitive leap critical points to obtain a learning development trajectory map. Step S4: Based on the learning development trajectory map, refine the learning patterns to obtain the learning style spectrum, and construct a cognitive sensitivity analysis model based on the learning style spectrum; Step S5: Intelligently combine learning navigation schemes through the cognitive sensitivity analysis model to obtain a set of personalized learning enhancement strategies; arrange teaching coordination schemes according to the set of personalized learning enhancement strategies to obtain a comprehensive education evaluation report, and provide learning guidance based on the comprehensive education evaluation report.

[0011] Preferably, step S1 specifically includes: Multimodal cognitive data and learning behavior trajectories are collected, and their structures are standardized to form a multi-source educational data spectrum. The multi-source educational data spectrum is then calibrated with time anchors to obtain time-consistent cognitive records and time-consistent behavioral trajectories. Based on these time-consistent records and trajectories, learning context mapping analysis is performed to identify contextual interaction relationships between data, thereby extracting core learning context elements. Based on these core learning context elements, intelligent context matching is performed to obtain a context fusion transformation rule chain. Finally, the time-consistent cognitive records are subjected to multi-source coupling transformation according to the context fusion transformation rule chain to obtain an educational interaction information spectrum.

[0012] Specifically, collecting multimodal cognitive data and learning behavior trajectories refers to collecting data on learners at different cognitive levels and their behavioral data sequences during the learning process through various technical means. Multimodal cognitive data includes attention data, eye-tracking data, and emotion recognition data. Attention data is acquired through a facial expression recognition system, which employs a deep learning-based facial feature recognition model. This model uses a convolutional neural network structure, containing eight convolutional layers and three fully connected layers. The input is a sequence of facial images, and the output is an attention concentration score. Eye-tracking data is collected using professional eye-tracking equipment, recording the learner's gaze placement, jump trajectory, and fixation duration on the learning material. Emotion recognition data is acquired through a multimodal emotion recognition system, which integrates facial expressions, speech features, and physiological signals to construct an emotion state assessment model.

[0013] The learning behavior trajectory includes the learner's operation sequence on the online learning platform, knowledge point access path, question-and-answer process, and resource usage pattern. The operation sequence records the learner's clicks, swipes, pauses, and other actions on the platform, along with their timestamps and contextual information. The knowledge point access path records the order in which the learner browses knowledge content and the time spent on each page. The question-and-answer process records the learner's thinking time, number of answer modifications, and final result. The resource usage pattern records the learner's preferences for different types of learning resources and their usage efficiency.

[0014] Structural standardization of multimodal cognitive data and learning behavior trajectories refers to converting data from different sources and in different formats into a unified data structure. The standardization process employs a data mapping framework, which includes a field mapping module, a data type conversion module, and an integrity verification module. The field mapping module maps field names from different data sources to a unified field definition system. The data type conversion module unifies the format of different data types, including the normalization of numerical, categorical, and text data. The integrity verification module checks the integrity of the data and intelligently imputes missing values ​​using methods such as mean imputation, mode imputation, and machine learning prediction imputation. The standardized data is stored in a structured tabular format, forming an educational multi-source data spectrum, including fields such as learner ID, timestamp, cognitive state indicators, behavioral operation type, learning content identifier, and environmental parameters.

[0015] Time anchor calibration of multi-source educational data refers to aligning data from different acquisition frequencies and time bases onto a unified time axis. The calibration process first identifies key event points in each data stream as time anchors, then performs interpolation and resampling based on these anchors. Key event point identification employs change point detection, locating important time nodes by monitoring changes in the statistical characteristics of the data sequence. Change point detection utilizes a Bayesian online change point detection algorithm, which adaptively identifies abrupt changes in the data stream. Data resampling employs a cubic spline interpolation method to ensure the smoothness and continuity of the data during the alignment process. Through time anchor calibration, time-consistent cognitive records and time-consistent behavioral trajectories with a unified time standard are obtained.

[0016] Learning context mapping analysis based on time-consistent cognitive records and time-consistent behavioral trajectories refers to analyzing the interaction between learning context factors and cognitive states. This process first involves extracting fluctuation features from both time-consistent cognitive records and time-consistent behavioral trajectories to obtain purified cognitive and behavioral data streams. Fluctuation feature extraction employs wavelet decomposition, a technique that decomposes the original signal into different frequency components, separating trend and noise components. Wavelet decomposition uses discrete wavelet transform with Daubechies wavelet basis functions, achieving a five-level decomposition level, effectively capturing the multi-scale characteristics of the signal. Wavelet denoising removes random noise and anomalous fluctuations, resulting in a smoother data stream.

[0017] Extracting cognitive development patterns from purified cognitive data streams refers to identifying patterns that reflect the learner's cognitive development patterns from the cognitive data. Pattern extraction employs a deep temporal feature learning network, which combines a Long Short-Term Memory (LSTM) network and an attention mechanism to capture long-term dependencies and key points of change in cognitive data. The network architecture includes a bidirectional LSTM layer, a temporal attention layer, and a feature fusion layer. The bidirectional LSTM layer, composed of 128 neurons, captures forward and backward dependencies in the sequence; the temporal attention layer calculates the importance weights of each time point in the sequence, highlighting cognitive features at critical moments; the feature fusion layer combines the LSTM output and attention weights to generate a comprehensive feature representation. Through this network, key learning inflection points in the cognitive process are identified, marking significant changes in cognitive level and obtaining a cognitive change feature map.

[0018] Identifying learning environment fluctuations in purification behavior data streams involves analyzing the changing patterns of learning environment factors and their impact on learning behavior. Environmental fluctuation identification employs a Conditional Random Field (CRF) model, which captures contextual dependencies in sequence data and identifies transition points in the environment. The CRF model's characteristic functions include state characteristic functions and transition characteristic functions. State characteristic functions describe the environmental characteristics at a single time point, while transition characteristic functions describe the environmental changes between adjacent time points. Model training utilizes maximum likelihood estimation, iteratively solving for the model parameters. Through the CRF model, significant changes in the learning environment and their characteristics are identified, resulting in an environmental impact feature map.

[0019] Pearson correlation analysis was conducted using cognitive change feature maps and environmental impact feature maps to obtain data on the intensity of cognitive-environment interaction. Pearson correlation analysis is a statistical method for measuring the degree of linear correlation between two variables, quantifying their correlation by calculating the correlation coefficient. During the analysis, key indicators from the cognitive change feature map were paired with environmental factors from the environmental impact feature map, and the correlation coefficient between each pair of factors was calculated. The correlation analysis results formed a correlation matrix, where each element represents the correlation strength of a pair of cognitive-environment factors. The correlation coefficient ranges from -1 to 1, with a larger absolute value indicating a stronger correlation. Significantly correlated factor pairs were filtered by setting correlation coefficient thresholds to obtain data on the intensity of cognitive-environment interaction.

[0020] This study analyzes the influencing factors of learning based on cognitive-environment interaction intensity data, employing structural equation modeling (SEM) to analyze the causal network between cognitive and environmental factors. The SEM consists of two parts: a measurement model and a structural model. The measurement model describes the relationship between latent and observed variables, while the structural model describes the causal relationships among latent variables. Model construction utilizes a combination of theoretical and data-driven approaches. First, an initial model structure is built based on educational psychology theories, and then the model is optimized and adjusted through model fit testing. Maximum likelihood estimation is used for model parameter estimation, and model evaluation metrics include chi-square value, Comparative Fit Index (CFI), and Root Mean Square Error of Approximation (RMSEA). Through SEM analysis, key factors with significant impact on cognitive development are identified, resulting in a set of cognitive moderating factors.

[0021] Contextual conditional sequence deconstruction of the cognitive modulatory factor set refers to analyzing the dynamic change patterns of these factors under different contextual conditions. Sequence deconstruction employs a Hidden Markov Model (HMM), which describes the transition process of the system state over time and the relationship between observed variables and hidden states. The model includes a state space, a state transition probability matrix, and an observation probability matrix. The state space is defined as different learning context types; the state transition probability matrix describes the transition patterns between context types; and the observation probability matrix describes the probability distribution of different cognitive modulatory factors observed under a specific context. By using the HMM, the dynamic change patterns of cognitive modulatory factors under different contextual conditions are deconstructed, obtaining a dynamic map of cognitive-environment interaction.

[0022] The magnitude of influence is calculated based on a dynamic graph of cognitive environment interactions, and the information gain method is used to quantify the degree of influence of different contextual factors on cognitive development. Information gain calculation is based on the principle of information entropy, assessing the importance of a factor by comparing the change in system entropy with and without the presence or absence of that factor. In the calculation process, the learner's cognitive state is used as the target variable, and the contextual factors are used as feature variables to calculate the information gain value for each contextual factor. A higher information gain value indicates a more significant impact of the contextual factor on the cognitive state. By setting an information gain threshold, key contextual factors with significant influence are selected to form core learning contextual elements.

[0023] Context-based intelligent matching, based on core learning context elements, refers to automatically matching the most suitable data fusion strategy according to the characteristics of the learning context. Context-based intelligent matching employs a knowledge graph-based reasoning system, which comprises three core components: a context ontology model, a rule base, and an inference engine. The context ontology model uses OWL to describe context concepts and their relationships, constructing a semantic network of contextual knowledge. The rule base contains a set of rules for context adaptation, represented using SWRL. The inference engine, based on a forward-linked reasoning mechanism, automatically infers from known contextual facts and the rule base to arrive at an adaptation conclusion. The matching process first maps core learning context elements to the context ontology model, then triggers relevant rules for reasoning, ultimately generating a context adaptation scheme. Through context-based intelligent matching, data fusion strategies for different contexts are obtained, forming a context fusion transformation rule chain.

[0024] Multi-source coupling transformation of temporally consistent cognitive records based on contextual fusion transformation rule chains refers to fusing and integrating cognitive data from different sources according to contextual adaptation rules. This multi-source coupling transformation employs a deep multimodal fusion network, capable of handling feature extraction and interactive fusion of data from different modalities. The network architecture includes a modality-specific encoder, a cross-modal attention module, and a multi-layer fusion decoder. The modality-specific encoder designs a dedicated feature extraction network for each data modality, such as a Transformer encoder for text data and a convolutional neural network encoder for image data. The cross-modal attention module calculates the correlation weights between features from different modalities to achieve information interaction between modalities. The multi-layer fusion decoder gradually integrates information from various modalities through a cascaded approach to generate a unified representation. Network training adopts an end-to-end approach, using a multi-task learning framework to simultaneously optimize feature extraction and fusion performance. Through the deep multimodal fusion network, multi-source data in temporally consistent cognitive records are contextually integrated to ultimately obtain an educational interaction information spectrum. This spectrum is a multidimensional dataset that integrates cognitive states, behavioral trajectories, and contextual association information, providing foundational data support for subsequent cognitive decoding.

[0025] Preferably, step S2 specifically includes: This process involves deconstructing the educational interaction information spectrum into a cognitive hierarchy, analyzing the relationships between knowledge units, and obtaining a cognitive structure hierarchy diagram. Based on this diagram, subject knowledge domains are mapped to obtain a learning behavior cognitive map. Cognitive depth is then measured to obtain a cognitive intensity distribution matrix. Knowledge associations are extracted from the learning behavior cognitive map to obtain cognitive node connection data. Cognitive patterns are combined using the cognitive intensity distribution matrix and the cognitive node connection data to obtain a cognitive pattern cluster diagram. Learning contexts are associated and classified using the cognitive pattern cluster diagram to obtain a contextual cognitive association report. Finally, a knowledge network topology is constructed based on the cognitive pattern cluster diagram and the contextual cognitive association report to obtain a cognitive development topography map.

[0026] Specifically, the cognitive hierarchy deconstruction of the educational interaction information spectrum refers to analyzing the hierarchical relationships of cognitive structures inherent in educational interaction data. The deconstruction process employs a knowledge representation learning model, which can extract knowledge concepts and their hierarchical relationships from educational interaction data. The model adopts a hierarchical representation learning architecture, including a concept encoding layer, a relationship inference layer, and a hierarchy construction layer. The concept encoding layer uses the pre-trained language model BERT as its foundation, fine-tuning it to adapt to concept representation tasks in the educational domain. The BERT model uses a 12-layer Transformer encoder architecture with 768 hidden layers, 12 attention heads, and approximately 110 million parameters. The relationship inference layer is implemented based on a graph attention network (GAT), which can automatically learn important relationships between nodes through an attention mechanism. The GAT network contains three graph attention layers, each with eight attention heads, and an output dimension of 64. The hierarchy construction layer uses a combination of rule-based and statistical methods to construct a hierarchical structure through hierarchical relationships, inclusion relationships, and prior-successor relationships between concepts. By deconstructing the cognitive hierarchy, we obtain a cognitive structure hierarchy diagram that represents the hierarchical relationship between knowledge concepts.

[0027] Mapping subject knowledge domains based on cognitive structure hierarchy graphs refers to the correspondence analysis between cognitive structures and the knowledge systems of specific disciplines. The mapping process employs domain knowledge graph alignment technology, which achieves the mapping between different knowledge systems through two steps: entity alignment and relation alignment. Entity alignment combines Bidirectional Encoding Representation Network (BERT) with knowledge graph embedding. First, BERT extracts the semantic features of concepts; then, the TransE model is used to learn the structural features of the knowledge graph; finally, cosine similarity is used to calculate the matching degree between concepts. Relation alignment determines the correspondence of relationships by comparing the similarity of paths between entities in two knowledge graphs. The mapping result forms a learning behavior cognitive graph, which is a knowledge network structure where nodes represent knowledge concepts, edges represent relationships between concepts, and it also contains information about the association between learning behaviors and knowledge concepts.

[0028] Cognitive depth measurement of a learning behavior cognitive map refers to assessing the learner's depth of mastery and understanding of various knowledge concepts. The measurement process employs a cognitive diagnostic model, which infers the learner's underlying cognitive state based on their performance. The model utilizes a deep cognitive diagnostic network architecture, integrating the representational power of neural networks with the interpretability of cognitive diagnostic models. The network structure includes an input layer, an embedding layer, an interaction layer, and an output layer. The input layer receives the learner's behavioral data; the embedding layer transforms discrete behavioral features into continuous vector representations; the interaction layer models the dependencies between knowledge concepts through an attention mechanism; and the output layer predicts the learner's mastery level for each knowledge point. Model training employs the expectation-maximization algorithm, iteratively optimizing model parameters and underlying cognitive states. Through cognitive depth measurement, a cognitive intensity distribution matrix representing the learner's mastery level of each knowledge concept is obtained.

[0029] Knowledge association extraction based on cognitive graphs of learning behavior refers to analyzing the semantic and structural relationships between knowledge concepts. Association extraction employs graph neural network (GCN) technology, which effectively handles node relationships in graph-structured data. The model uses a Graph Convolutional Network (GCN) architecture, capturing local structural information of nodes through a message-passing mechanism. The GCN model contains three graph convolutional layers, with output dimensions of 128, 64, and 32 for each layer, using ReLU activation functions. The final layer uses a sigmoid activation function to output the association strength between nodes. Model training employs an edge prediction task, using known knowledge associations as training samples to learn the ability to predict unknown associations. Through the graph neural network, explicit and implicit associations between knowledge concepts are extracted, obtaining cognitive node connection data.

[0030] Cognitive patterns are combined based on the cognitive intensity distribution matrix and cognitive node connection data. First, knowledge mastery is analyzed based on the cognitive intensity distribution matrix, and a knowledge state space model is used to describe the learners' knowledge mastery distribution characteristics. The model is based on multidimensional item response theory, assuming that learners' knowledge mastery status can be represented by multiple latent trait dimensions. The model includes item difficulty parameters, discrimination parameters, and guessing parameters, and the model parameters are estimated using the maximum marginal likelihood estimation method. Based on the estimated parameters, the learners' positional distribution in the knowledge space is calculated, obtaining a knowledge mastery distribution map. The knowledge mastery distribution map is then divided into ability levels, and a hierarchical clustering algorithm is used to identify knowledge regions at different mastery levels. The clustering algorithm uses Ward's minimum variance method, classifying knowledge points into different ability levels based on the similarity of knowledge mastery, obtaining cognitive ability level data.

[0031] Based on cognitive ability level data, cognitive weakness areas are identified, and anomaly detection technology is used to identify weak links in knowledge mastery. Anomaly detection constructs a set of random decision trees and calculates the isolation degree of samples to determine whether they are outliers. The algorithm parameters are set to 100 trees and a subsampling size of 256. By calculating isolation scores, knowledge areas with significantly low cognitive abilities are identified, thus obtaining cognitive weakness data. Knowledge association discontinuities are identified on cognitive node connection data, using community detection and breakpoint analysis techniques to identify structural discontinuities in the knowledge network. Community detection discovers the community structure in the knowledge network by optimizing the modularity index; breakpoint analysis identifies weak connections between communities by calculating edge betweenness and node betweenness. Through comprehensive analysis, regions with significant fluctuations in knowledge association strength are identified, obtaining data on regions with fluctuating knowledge association strength.

[0032] Knowledge domain interaction analysis was conducted on cognitive gap data and knowledge association fluctuation area data, employing graph-based association rule mining techniques to discover interaction patterns between the two types of data. The analysis process first constructs a knowledge domain interaction graph, where nodes represent knowledge points and edges represent the interaction relationships between knowledge points. Then, based on frequent subgraph mining, frequently occurring substructure patterns are discovered. Finally, through importance assessment of subgraph patterns, patterns with high support and confidence are selected as key interaction patterns. Through knowledge domain interaction analysis, key influencing areas of cognitive development are identified, obtaining key cognitive influencing area data. Cognitive pattern clustering is performed based on this key cognitive influencing area data, using a spectral clustering algorithm to group cognitive patterns. Spectral clustering first constructs a similarity matrix, then calculates the Laplace transform of the matrix, extracts feature vectors, and finally applies the K-means algorithm to cluster in the feature space. The number of clusters is determined using evaluation metrics such as silhouette coefficient. Through cognitive pattern clustering, a cognitive pattern cluster graph is obtained, which shows the distribution and association structure of different cognitive patterns.

[0033] Learning context association and classification are performed using cognitive pattern clustering graphs. First, boundary detection of learning contexts is conducted based on the cognitive pattern clustering graphs, employing boundary detection algorithms to identify boundary regions of different learning contexts. Boundary detection utilizes density peak clustering, which identifies cluster centers and boundary points by calculating local density and distance metrics. The algorithm first calculates the local density of each point and the minimum distance to high-density points, then identifies cluster centers and boundary points using a density-distance map. Boundary detection yields learning context domain range data describing the scope of different learning contexts. Cognitive associations within learning contexts are analyzed based on this domain range data, using association analysis techniques to uncover association patterns between cognitive elements in different contexts. Association analysis filters significant association rules by setting minimum support and minimum confidence thresholds. Through association analysis, a cognitive association network is constructed for different learning contexts, resulting in a cognitive localization map of the learning context.

[0034] Learning style traits are identified from cognitive pattern cluster graphs using feature selection and pattern recognition techniques. Feature selection employs Recursive Feature Elimination (RFE), which repeatedly trains a classifier and removes the least important features to select the most representative learning style traits. The classifier uses a random forest model with 100 decision trees and a maximum depth of 10, selecting the top 20% of key features based on average feature importance scores. Through feature selection, learning style trait data representing the main characteristics of learning styles are obtained. Based on this data, learning style types are classified using unsupervised learning methods. This classification yields a learning style type spectrum describing different learning style types and their characteristics.

[0035] Cognitive adaptability is assessed based on learning style trait data, employing a reinforcement learning framework to evaluate learners' adaptability in different learning environments. The assessment model treats the learning process as a Markov decision process, with the state space representing learning context features, the action space representing learning strategy selection, and the reward function representing learning performance scores. The model uses a deep Q-network structure containing three fully connected layers with hidden layer dimensions [128, 64]. The input is a combined representation of the learning context and learner features, and the output is a value assessment of different learning strategies. Through model training and evaluation, cognitive adaptability data describing learners' adaptability in different contexts is obtained. The learning context cognitive localization map, learning style type spectrum, and cognitive adaptability data are integrated to incorporate learning context characteristics, using multi-view learning technology to fuse contextual feature information from different perspectives. The integration model uses tensor decomposition, treating the three types of data as different views of the same tensor, and extracting shared and unique information using the Tucker decomposition method. Model parameters are optimized using alternating least squares, and the loss function includes a reconstruction error term and a regularization term. Through multi-view fusion, a comprehensive contextual cognitive association report describing the characteristics of the learning context is obtained.

[0036] A knowledge network topology was constructed based on cognitive pattern cluster graphs and contextual cognitive association reports, employing graph structure optimization techniques to build a network representation of cognitive development. The construction process included three steps: node definition, edge relationship establishment, and graph structure optimization. Node definition was based on cognitive pattern clusters and knowledge concepts, with each node containing information on cognitive level, knowledge attributes, and learning style. Edge relationships were established based on knowledge dependencies and learning path relationships, with edge weights calculated using cognitive association strength and learning frequency. Graph structure optimization simulated physical attraction and repulsion to achieve visual balance in the graph structure. Furthermore, the topology construction showcased the multi-layered organizational characteristics of the cognitive network through a hierarchical community structure. The final constructed cognitive development topography is a multi-dimensional network structure that not only demonstrates the relationships between knowledge concepts but also includes fluctuations in cognitive levels and diverse learning path choices, providing a foundation for subsequent cognitive discontinuity detection.

[0037] Preferably, step S3 specifically includes: This process involves collecting real-time teaching interaction sequences and extracting key information from them to obtain the core data stream of teaching interaction. Cognitive leap points are identified based on a cognitive development topography map to obtain a cognitive leap pattern map. Cognitive fault types are categorized from the cognitive leap pattern map to obtain a set of cognitive leap critical points. Key learning nodes are selected from the core data stream of teaching interaction based on the set of cognitive leap critical points to obtain high-value interaction data. Learning path nodes are identified from this high-value interaction data to obtain a sequence of cognitive evolution nodes. The learning development quality of the cognitive evolution node sequence is evaluated based on the high-value interaction data to obtain a cognitive node quality index. Finally, cognitive evolution is tracked based on the cognitive evolution node sequence and the cognitive node quality index to obtain a learning development trajectory map.

[0038] Specifically, collecting real-time teaching interaction sequences refers to the real-time collection of data sequences related to teacher-student interactions and student learning processes during the teaching process. Data collection is conducted through multiple channels, including smart classroom systems, online learning platforms, and mobile learning applications. Smart classroom systems incorporate multimodal sensor networks capable of capturing audio, video, and interactive information in the classroom environment. Online learning platforms record learners' operational behaviors, response times, and interaction patterns through embedded data acquisition modules. Mobile learning applications collect learners' learning behavior data on mobile devices via client SDKs. Data collection employs an event-driven architecture, building a real-time data stream processing pipeline based on Apache Kafka to ensure low-latency data transmission and processing.

[0039] This study extracts key information from real-time interactive teaching sequences, employing natural language processing and multimodal analysis techniques to extract core information from the interactive data. Text information extraction utilizes a BERT-based sequence labeling model, capable of identifying key entities and events within the text. The model architecture includes a BERT encoding layer and a Conditional Random Field (CRF) output layer. The BERT model is pre-trained and fine-tuned to adapt to specific educational tasks. Speech information extraction employs a deep speech recognition model based on a Transformer-Transducer architecture, comprising an encoder, a prediction network, and a connection network. Video information extraction utilizes a 3D convolutional neural network (3D-CNN) to extract spatiotemporal features from video sequences. Multimodal information fusion employs an attention-based multimodal Transformer model, achieving interaction and integration of different modalities through cross-modal attention layers. Through key information extraction, a core data stream of interactive teaching data, including teaching focus, learning difficulties, and interactive focal points, is obtained.

[0040] This study identifies cognitive transition points based on a cognitive development topographic map, employing graph structure analysis and temporal pattern mining techniques to pinpoint key shifts in the cognitive development process. The identification process begins by using graph analysis algorithms to calculate node centrality and structural hole indices in the cognitive development topographic map, identifying key nodes and bridging nodes in the network. Node centrality calculation utilizes multiple centrality indices, including degree centrality, betweenness centrality, and eigenvector centrality, and comprehensively evaluates node importance through a weighted combination. Structural hole identification uses constraint coefficients and structural hole efficiency indices to discover information bridging locations within the knowledge network. Temporal pattern mining employs sequence pattern mining algorithms to discover frequent temporal patterns in the cognitive development process. Through comprehensive analysis, key transition points in the cognitive development process are identified, resulting in a cognitive transition pattern map.

[0041] This paper summarizes the cognitive discontinuity types in a cognitive leap pattern map, employing pattern recognition and type learning methods to analyze the types of cognitive leap points. Type summarization utilizes a semi-supervised learning framework, combining expert knowledge and data-driven methods. The model architecture combines graph representation learning and cluster analysis. First, structural and semantic features of leap points are extracted using a graph neural network. Then, hierarchical clustering algorithms are used to group the feature vectors. The graph neural network uses a graph attention network (GAT) structure, which adaptively aggregates neighborhood information through an attention mechanism. Expert knowledge guides the clustering process with a small number of labeled samples, improving the accuracy and interpretability of classification. Through type summarization, cognitive leap points are classified into four main types: knowledge discontinuity, understanding discontinuity, application discontinuity, and innovation discontinuity, forming a set of cognitive leap critical points.

[0042] Based on the set of cognitive leap thresholds, key learning nodes are screened from the core data stream of teaching interactions. A correlation-based filtering method is used to identify interactive events related to cognitive leaps. The screening process first calculates the correlation between interactive events and cognitive leap points, based on three dimensions: temporal approximation, content similarity, and contextual relevance. Temporal approximation is calculated using a time window function, content similarity using semantic similarity, and contextual relevance using graph path analysis. A weighted combination approach is used for correlation calculation, with the weights of the three dimensions determined through cross-validation. By setting a correlation threshold, interactive events highly correlated with cognitive leaps are selected, forming high-value interactive data. Learning path nodes are identified from the high-value interactive data, using sequence pattern mining and key event extraction techniques to identify key nodes in the learning path. Sequence pattern mining uses an improved SPADE algorithm, which discovers frequent sequence patterns in the learning process by setting minimum support and maximum interval constraints. Key event extraction is based on event impact assessment, determining the importance of an event by calculating its impact on subsequent learning processes. Through node identification, a sequence of cognitive evolution nodes representing key stages of cognitive development is obtained.

[0043] The learning development quality of cognitive evolution node sequences is assessed based on high-value interactive data, employing a multi-dimensional evaluation method to measure the quality characteristics of cognitive development. The evaluation dimensions include cognitive depth, knowledge connectivity, and learning stability. Cognitive depth is assessed using a deep learning network based on Bloom's Taxonomy of Cognition, inferring cognitive levels by analyzing learning behavior and performance. Knowledge connectivity is assessed using knowledge graph analysis, evaluating the degree of knowledge integration by calculating the connection strength and structural characteristics between knowledge points. Learning stability is assessed using time series analysis, evaluating the stability of the learning process by calculating the volatility and trend of learning performance. The evaluation results are presented using a comprehensive scoring method, with a weighted average calculated to obtain a quality index for each cognitive node, forming a cognitive node quality index.

[0044] Cognitive evolution is tracked based on a sequence of cognitive evolution nodes and a cognitive node quality index. The tracking process first constructs a cognitive state transition network, where nodes represent cognitive states, edges represent transitions between states, and edge weights are calculated based on a combination of transition difficulty and quality gains. Path planning employs the A* algorithm, optimizing search efficiency through a defined heuristic function. This heuristic function considers both target distance and path quality, ensuring that the found paths are both efficient and high-quality. Through path planning, key nodes in the cognitive evolution node sequence are connected to form a complete learning development trajectory graph. This learning development trajectory graph is a multi-dimensional representation that not only shows the path and direction of cognitive development but also includes information such as development speed, quality, and key turning points, providing a comprehensive view for learning pattern analysis.

[0045] Preferably, step S4 specifically includes: This process involves extracting learning pattern characteristics from the learning development trajectory map to obtain a learning pattern feature set; analyzing cognitive processing tendencies based on the learning pattern feature set to obtain an initial cognitive tendency map; acquiring real-time learning behavior stream data and evaluating the cognitive efficacy of the real-time learning behavior stream data to obtain real-time cognitive efficacy data; analyzing cognitive processing tendencies based on real-time cognitive efficacy data to obtain an updated cognitive tendency map; performing cognitive style evolution analysis based on the initial and updated cognitive tendency maps to obtain a cognitive style evolution trajectory; extracting learning elasticity features based on the cognitive style evolution trajectory and real-time cognitive efficacy data to obtain a learning style spectrum; and constructing a cognitive sensitivity analysis model based on the learning style spectrum.

[0046] Specifically, the learning pattern characteristics in the learning development trajectory graph are extracted, and a combination of feature engineering and deep feature learning is used to extract multi-dimensional learning features. The feature extraction process includes two parts: manual feature design and automatic feature learning. Manual feature design is based on educational psychology theory, extracting key indicators related to learning styles, including learning speed features, rhythm change features, path selection features, and cognitive transition features. Automatic feature learning uses graph representation learning technology, learning the latent representation of the trajectory graph through a graph convolutional autoencoder. The graph convolutional autoencoder consists of an encoder and a decoder. The encoder contains three graph convolutional layers with hidden layer dimensions of [64, 32], and the decoder uses a graph transposed convolutional structure for graph reconstruction. Through feature extraction, a learning pattern feature set containing explicit and implicit features is obtained.

[0047] This study analyzes cognitive processing tendencies based on learning pattern feature sets and employs a cognitive style recognition model to analyze learners' preferences in cognitive processing methods. The cognitive style model, based on the Felder-Silverman learning style model, categorizes cognitive styles into four dimensions: perceptual-intuitive, visual-verbal, active-reflective, and sequential-holistic. The model utilizes a multi-label classification framework, capable of simultaneously predicting learners' style preferences across multiple dimensions. The model architecture employs a deep multi-task learning network, including a shared representation layer and task-specific layers. The shared representation layer consists of two fully connected layers, taking the learning pattern feature set as input and outputting a shared representation vector. The task-specific layers design a dedicated classification head for each style dimension, using a sigmoid activation function to output style preference probabilities. Model training employs a multi-task loss function, combining binary cross-entropy loss and label relevance regularization. Through model inference, an initial cognitive tendency map representing learners' cognitive processing methods is obtained.

[0048] The system acquires real-time learning behavior data, collecting learners' current learning behavior information through multiple channels. Data acquisition utilizes a multimodal perception system, including technologies such as action behavior recording, eye tracking, and facial expression recognition. Action behavior recording employs an event log system to record learners' interactive operations such as clicks, swipes, and inputs; eye tracking uses an infrared eye tracker to track learners' gaze movements and fixation patterns; facial expression recognition uses computer vision technology to analyze changes in learners' facial expressions. The data acquisition system adopts a distributed architecture, with a front-end acquisition module responsible for data acquisition, an edge computing module responsible for data preprocessing, and a cloud storage module responsible for data integration and persistence.

[0049] This study evaluates the cognitive efficacy of real-time learning behavior stream data, employing a real-time evaluation framework to analyze the cognitive value and learning effectiveness of learning behaviors. The evaluation model utilizes a deep reinforcement learning architecture, treating the learning process as a Markov decision process, with learning states as behavioral feature vectors, learning actions as cognitive operation types, and reward signals as learning effectiveness scores. The core model employs a Double-Q Network (DQN) structure, containing two identical neural networks: a main network for action selection and a target network for value function estimation. The network architecture includes three fully connected layers with hidden layer dimensions of [128, 64] and the activation function being LeakyReLU. Through model evaluation, cognitive efficacy scores for different learning behaviors are calculated, generating real-time cognitive efficacy data.

[0050] Based on real-time cognitive efficacy data analysis of cognitive processing tendencies, this study employs the same cognitive style model architecture as the initial analysis, but adjusts the model using transfer learning. The transfer learning process first initializes the parameters using the initial model, then fine-tunes the training using real-time data. Fine-tuning utilizes a small learning rate and a limited number of iterations to maintain model stability while adapting to the characteristics of the new data. Furthermore, a data importance weighting mechanism is introduced during fine-tuning, adjusting sample weights based on the temporal approximation and quality reliability of the data to ensure that the most recent and reliable data has a greater impact. Through model updates, an updated cognitive tendency map based on the latest learning behavior is obtained.

[0051] This study analyzes the evolution of cognitive style based on initial and updated cognitive tendency maps, employing temporal pattern mining and change point detection to investigate the dynamic characteristics of cognitive style changes. Temporal pattern mining utilizes a recurrent neural network (RNN) model to capture the evolutionary patterns of cognitive style over time. The input is cognitive style features in a time series, and the output is a style change prediction. Key moments of style transition are identified by monitoring abrupt changes in the probability distribution. Through comprehensive analysis, a cognitive style evolution trajectory describing the evolutionary characteristics of cognitive style is obtained.

[0052] Learning resilience features are extracted based on cognitive style evolution trajectories and real-time cognitive efficacy data. Adaptive learning analytics is employed to study learners' adaptive characteristics in different contexts. Feature extraction first constructs a learning environment-style-efficacy relationship model, using structural equation modeling to describe the causal network among the three. The structural equation model includes a measurement model and a structural model, and the model parameters are estimated using maximum likelihood estimation. Based on model analysis, features such as learners' adaptation speed, adaptation magnitude, and stability under environmental changes are extracted. Furthermore, frequency domain analysis is used in feature extraction, transforming the learning behavior sequence to the frequency domain through wavelet transform to analyze the energy distribution characteristics of different frequency components. Through multi-dimensional feature extraction, a learning style spectrum describing the dynamic characteristics of learning styles is obtained.

[0053] A cognitive sensitivity analysis model is constructed based on the learning style spectrum. The model architecture integrates three technologies: knowledge graph, deep learning, and rule-based reasoning, forming a hybrid intelligent system with reasoning and adaptive capabilities. The knowledge graph component uses an RDF triple structure to store cognitive style knowledge, including concepts and relationships such as style type, feature representation, and adaptation strategies. The deep learning component employs graph neural networks and attention mechanisms to achieve style recognition and sensitivity analysis through end-to-end learning. The rule-based reasoning component uses a forward-linked reasoning mechanism to generate teaching suggestions based on predefined rules and derived facts. Model training adopts a multi-stage approach: first, each component is pre-trained, and then end-to-end joint optimization is performed. Through model construction, a cognitive sensitivity analysis model capable of analyzing learners' cognitive sensitivity characteristics and generating personalized learning suggestions is formed.

[0054] Preferably, step S5 specifically includes: This process involves identifying and updating unconventional learning patterns in a cognitive tendency map to obtain data on cognitive anomaly patterns. A cognitive sensitivity analysis model is used to intelligently combine these patterns into a set of personalized learning enhancement strategies. Based on this set of strategies and a cognitive development topography map, learning intervention schemes are simulated to obtain data predicting the effectiveness of these interventions. This data is then used to coordinate multi-party teaching resources to obtain integrated teaching resource data. Finally, adaptive strategies are applied to the integrated teaching resource data based on the cognitive transition pattern map to generate a comprehensive educational evaluation report. This report is then used to provide learning guidance.

[0055] Specifically, this study identifies and updates unconventional learning patterns in the cognitive tendency map, employing multiple anomaly detection techniques to identify anomalous features within these patterns. Anomaly detection combines three approaches: statistical methods, distance methods, and density methods. The statistical method utilizes a modified Z-score algorithm, identifying statistical outliers by calculating the standardized distance between feature values ​​and the mean. The distance method detects outliers by comparing the local density differences between samples and their neighborhoods. The density method employs the DBSCAN clustering algorithm, identifying anomalous samples that do not belong to any dense region by recognizing noise points. The anomaly detection results are integrated using an ensemble learning method, with a majority voting mechanism determining the final anomaly determination. Through anomaly detection, unconventional patterns in cognitive tendencies are identified, yielding cognitive anomaly pattern data.

[0056] This study employs a cognitive sensitivity analysis model to intelligently combine learning navigation schemes for cognitive anomaly pattern data, generating personalized learning schemes using a hybrid decision-making system based on knowledge reasoning and deep reinforcement learning. The scheme generation process comprises three stages: needs analysis, strategy retrieval, and scheme combination. In the needs analysis stage, cognitive anomaly patterns are transformed into learning needs descriptions, and key needs features are extracted using natural language processing techniques. In the strategy retrieval stage, strategy components related to the learning needs are retrieved based on a knowledge graph, and candidate strategies are screened using semantic similarity and association rule matching. In the scheme combination stage, deep reinforcement learning methods are used to optimize strategy combination, modeling the strategy combination problem as a sequential decision problem, and learning the optimal strategy combination order through a deep Q-network. The reinforcement learning model adopts a hierarchical reinforcement learning architecture, including two decision levels: high-level strategy selection and low-level parameter adjustment. Through intelligent combination of learning navigation schemes, a personalized learning enhancement strategy set for specific cognitive anomaly patterns is obtained.

[0057] This simulation method simulates learning intervention programs based on personalized learning enhancement strategy sets and cognitive development topography, and uses educational scenario simulation technology to predict the effectiveness of these interventions. The simulation system employs an agent-based educational ecosystem simulation framework, including three roles: learner agent, teaching agent, and environment agent. The learner agent, built on a cognitive psychology model, simulates learning behaviors with different cognitive styles and ability levels; the teaching agent, built on a teaching strategy library, implements teaching interventions based on learning enhancement strategies; and the environment agent, built on a cognitive development topography, provides a simulated environment for learning situations and knowledge structures. The simulation process uses a discrete event simulation method, setting different combinations of intervention strategies and observing the cognitive development trajectory of the learner agent. The simulation experiment uses a controlled experimental design to compare the differences in learning effects under different intervention strategies. Through the simulation of learning intervention programs, the expected effects of different strategy combinations are evaluated, forming predictive data for learning intervention effects.

[0058] This study coordinates multi-party teaching resources based on learning intervention effect prediction data, employing a multi-objective optimization method to balance the allocation and coordination of different teaching resources. The resource coordination process first identifies available teaching resource types, including human resources (teachers, teaching assistants, peers), technological resources (learning platforms, educational applications, intelligent assistants), and content resources (textbooks, videos, exercises). Then, a resource-effect mapping model is established to describe the impact of different resource combinations on learning outcomes. The mapping model uses a deep learning-based multi-input multi-output network, with the input being a resource allocation vector and the output being a vector of expected learning outcomes. Based on the mapping model, a multi-objective optimization problem is constructed, with objective functions including maximizing learning outcomes, minimizing resource costs, and optimizing fairness. The optimization algorithm uses a non-dominated sorting genetic algorithm to find the Pareto optimal solution set through evolutionary computation. From the optimal solution set, the final resource allocation scheme is selected based on decision preferences, forming integrated teaching resource data.

[0059] Based on a cognitive transition pattern map, adaptive strategies are adjusted using integrated teaching resource data, and a dynamic adjustment mechanism is constructed using adaptive control theory. This mechanism, based on a feedback control system, comprises three modules: state monitoring, deviation analysis, and strategy adjustment. The state monitoring module acquires learning process state data through a real-time data acquisition system; the deviation analysis module calculates developmental deviations and transition risks by comparing actual states with expected trajectories; and the strategy adjustment module dynamically adjusts teaching strategies and resource allocation based on deviation and risk information. The current control strategy is optimized by predicting future state trajectories. Strategy adjustment pays particular attention to the learning process near cognitive transition points, helping learners successfully complete cognitive transitions through enhanced support and targeted interventions. Through adaptive strategy adjustment, a comprehensive educational evaluation report is generated, including educational objectives, implementation strategies, resource allocation, and assessment methods.

[0060] Learning guidance is based on a comprehensive educational evaluation report, employing a multi-level, multi-channel approach to implement personalized educational intervention. The guidance levels include cognitive, affective, and behavioral levels. Cognitive guidance targets knowledge comprehension and thinking patterns, using methods such as concept mapping, analogical examples, and guided questioning. Affective guidance addresses learning motivation and emotional state, employing methods such as positive feedback, emotional support, and goal setting. Behavioral guidance focuses on learning habits and skills, using methods such as demonstration, guided practice, and immediate feedback. Guidance channels include face-to-face instruction, online tutoring, and intelligent assistants. Face-to-face instruction is provided by teachers or tutors, offering personalized direct guidance; online tutoring is implemented through learning platforms and video conferencing tools, providing flexible remote guidance; and intelligent assistants are implemented through chatbots and intelligent recommendation systems, providing real-time automated guidance. Throughout the learning guidance process, a continuous improvement mechanism is employed, with regular assessments and feedback adjustments to continuously optimize guidance strategies and enhance educational effectiveness.

[0061] This embodiment achieves comprehensive data integration of learners' cognitive processes through the collection and multi-source coupling of multimodal cognitive data and learning behavior trajectories. It transforms scattered learning behavior data into a systematic educational interaction information spectrum, thereby enhancing the comprehensiveness of educational evaluation. Through cognitive structure decoding and knowledge network topology construction, the system can accurately map learners' cognitive development, providing a three-dimensional presentation of their cognitive structures and overcoming the limitations of the one-sidedness and subjectivity of traditional educational evaluation methods. By detecting cognitive fault lines and identifying sets of cognitive transition critical points, it can keenly capture key nodes in learners' cognitive development, track the cognitive evolution of real-time teaching interaction sequences, and generate learning development trajectory maps. This dynamic tracking mechanism significantly improves the timeliness and accuracy of educational evaluation, avoiding the static and lagging problems of traditional evaluation methods. Through learning pattern purification and learning style spectrum analysis, the system can comprehensively perceive learners' cognitive characteristics and learning preferences, constructing personalized cognitive sensitivity analysis models to effectively address the diverse needs of different learners and improve the adaptability and inclusiveness of the evaluation system. By introducing a cognitive sensitivity analysis model to intelligently combine learning navigation schemes, the system can analyze learners' cognitive states in real time, predict potential learning obstacles, and dynamically adjust teaching resources and strategies through the generation of personalized learning enhancement strategy sets, ensuring accurate learning guidance even when learning scenarios change. Based on the personalized learning enhancement strategy sets, the system coordinates teaching schemes and automatically generates comprehensive evaluation reports based on real-time learning status, guiding teaching practice and improving the educational evaluation system's ability to cope with complex learning situations, significantly enhancing learning efficiency and teaching quality. Example 2

[0062] Please see Figure 2 As shown, parts not described in detail in this embodiment are described in Embodiment 1. A system for comprehensive educational evaluation based on a smart education AI model is provided, including: Data fusion module: Collects multimodal cognitive data and learning behavior trajectories, and performs multi-source coupling of multimodal cognitive data and learning behavior trajectories through contextual intelligence integration to obtain the educational interaction information spectrum; Cognitive Topology Construction Module: Decodes the cognitive structure of the educational interaction information spectrum to obtain a cognitive graph of learning behavior, constructs the knowledge network topology based on the cognitive graph of learning behavior, and then obtains a cognitive development topography map; Cognitive Tracking Module: Collects real-time teaching interaction sequences; performs cognitive fault detection on the cognitive development topography map to obtain a set of cognitive leap critical points, and tracks the cognitive evolution of real-time teaching interaction sequences based on the set of cognitive leap critical points to obtain a learning development trajectory map; Model building module: Based on the learning development trajectory map, learning patterns are purified to obtain the learning style spectrum, and a cognitive sensitivity analysis model is built based on the learning style spectrum; Evaluation and guidance module: Intelligent combination of learning navigation schemes through cognitive sensitivity analysis model to obtain a set of personalized learning enhancement strategies; arrangement of teaching coordination schemes based on the set of personalized learning enhancement strategies to obtain a comprehensive education evaluation report; and learning guidance based on the comprehensive education evaluation report. The modules are connected via wired and / or wireless means to enable data transmission between them.

[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0064] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0065] In the description of this invention, it should be understood that the terms "first," "second," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0066] In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0067] In the description of this invention, "several" means one or more, and "a large number" means two or more.

[0068] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0069] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0070] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for comprehensive educational evaluation based on a smart education AI model, characterized in that, include: Step S1: Collect multimodal cognitive data and learning behavior trajectories, and perform multi-source coupling on the multimodal cognitive data and learning behavior trajectories to obtain the educational interaction information spectrum; Step S2: Decode the cognitive structure of the educational interaction information spectrum to obtain a cognitive map of learning behavior, construct the knowledge network topology based on the cognitive map of learning behavior, and then obtain a cognitive development topography map; Step S3: Collect real-time teaching interaction sequences; Cognitive fault detection is performed on the cognitive development topography to obtain a set of cognitive transition critical points, and cognitive evolution is tracked on the real-time teaching interaction sequence based on the set of cognitive transition critical points to obtain a learning development trajectory map. Step S4: Based on the learning development trajectory map, refine the learning patterns to obtain the learning style spectrum, and construct a cognitive sensitivity analysis model based on the learning style spectrum; Step S5: Intelligently combine learning navigation schemes through the cognitive sensitivity analysis model to obtain a set of personalized learning enhancement strategies; arrange teaching coordination schemes according to the set of personalized learning enhancement strategies to obtain a comprehensive education evaluation report, and provide learning guidance based on the comprehensive education evaluation report.

2. The method for comprehensive educational evaluation based on a smart education AI model according to claim 1, characterized in that, Step S1 includes: Multimodal cognitive data and learning behavior trajectories are collected, and their structures are standardized to form a multi-source educational data spectrum. The multi-source educational data spectrum is then calibrated with time anchors to obtain time-consistent cognitive records and time-consistent behavioral trajectories. Based on these time-consistent records and trajectories, learning context mapping analysis is performed to identify contextual interaction relationships between data, thereby extracting core learning context elements. Based on these core learning context elements, intelligent context matching is performed to obtain a context fusion transformation rule chain. Finally, the time-consistent cognitive records are subjected to multi-source coupling transformation according to the context fusion transformation rule chain to obtain an educational interaction information spectrum.

3. The method for comprehensive educational evaluation based on a smart education AI model according to claim 2, characterized in that, The learning context mapping analysis based on time-consistent cognitive records and time-consistent behavioral trajectories analyzes the contextual interaction relationships between data, and then extracts core learning context elements, including: Fluctuation features were extracted from time-consistent cognitive records and time-consistent behavioral trajectories to obtain purified cognitive data streams and purified behavioral data streams. Cognitive development patterns were extracted from the purified cognitive data streams to identify key learning inflection points and obtain a cognitive change feature map. Learning environment fluctuations were identified from the purified behavioral data streams to obtain an environmental impact feature map. Pearson correlation analysis was performed on the cognitive change feature map and the environmental impact feature map to obtain cognitive-environment interaction intensity data. Based on this data, learning influencing factors were analyzed to obtain a cognitive regulation factor set. The cognitive regulation factor set was deconstructed into a contextual condition sequence to obtain a cognitive-environment interaction dynamic map. The magnitude of influence was calculated based on the cognitive-environment interaction dynamic map to obtain core learning context elements.

4. The method for comprehensive educational evaluation based on a smart education AI model according to claim 1, characterized in that, Step S2 includes: This process involves deconstructing the educational interaction information spectrum into a cognitive hierarchy, analyzing the relationships between knowledge units, and obtaining a cognitive structure hierarchy diagram. Based on this diagram, subject knowledge domains are mapped to obtain a learning behavior cognitive map. Cognitive depth is then measured to obtain a cognitive intensity distribution matrix. Knowledge associations are extracted from the learning behavior cognitive map to obtain cognitive node connection data. Cognitive patterns are combined using the cognitive intensity distribution matrix and the cognitive node connection data to obtain a cognitive pattern cluster diagram. Learning contexts are associated and classified using the cognitive pattern cluster diagram to obtain a contextual cognitive association report. Finally, a knowledge network topology is constructed based on the cognitive pattern cluster diagram and the contextual cognitive association report to obtain a cognitive development topography map.

5. The method for comprehensive educational evaluation based on a smart education AI model according to claim 4, characterized in that, The process involves extracting knowledge associations based on a cognitive behavior graph to obtain cognitive node connection data, and combining cognitive patterns according to the cognitive intensity distribution matrix and the cognitive node connection data to obtain a cognitive pattern cluster graph, including: Knowledge mastery analysis is performed based on the cognitive intensity distribution matrix to obtain a knowledge mastery distribution map. This map is then divided into ability levels to obtain cognitive ability level data. Cognitive weakness areas are identified based on the cognitive ability level data to obtain cognitive weakness data. Knowledge association discontinuities are identified based on the cognitive node connection data to obtain knowledge association fluctuation area data. Knowledge domain interaction analysis is performed on the cognitive weakness data and the knowledge association fluctuation area data to obtain key cognitive influence area data. Cognitive pattern clustering is then performed based on the key cognitive influence area data to obtain a cognitive pattern cluster map.

6. The method for comprehensive educational evaluation based on a smart education AI model according to claim 4, characterized in that, The process of associating and classifying learning contexts through cognitive pattern clustering graphs to obtain a contextual cognitive association report includes: Learning context boundary detection is performed based on cognitive pattern cluster maps to obtain learning context domain range data. Cognitive associations within the learning context are analyzed based on this domain range data to obtain a learning context cognitive localization map. Learning style trait identification is performed on the cognitive pattern cluster maps to obtain learning style trait data, and learning style type classification is performed based on this data to obtain a learning style type spectrum. Cognitive adaptability assessment is performed based on the learning style trait data to obtain cognitive adaptability data. Finally, the learning context cognitive localization map, learning style type spectrum, and cognitive adaptability data are integrated to obtain a context-based cognitive association report.

7. The method for comprehensive educational evaluation based on a smart education AI model according to claim 1, characterized in that, Step S3 includes: This process involves collecting real-time teaching interaction sequences and extracting key information from them to obtain the core data stream of teaching interaction. Cognitive leap points are identified based on a cognitive development topography map to obtain a cognitive leap pattern map. Cognitive fault types are categorized from the cognitive leap pattern map to obtain a set of cognitive leap critical points. Key learning nodes are selected from the core data stream of teaching interaction based on the set of cognitive leap critical points to obtain high-value interaction data. Learning path nodes are identified from this high-value interaction data to obtain a sequence of cognitive evolution nodes. The learning development quality of the cognitive evolution node sequence is evaluated based on the high-value interaction data to obtain a cognitive node quality index. Finally, cognitive evolution is tracked based on the cognitive evolution node sequence and the cognitive node quality index to obtain a learning development trajectory map.

8. The method for comprehensive educational evaluation based on a smart education AI model according to claim 1, characterized in that, Step S4 includes: This process involves extracting learning pattern characteristics from the learning development trajectory map to obtain a learning pattern feature set; analyzing cognitive processing tendencies based on the learning pattern feature set to obtain an initial cognitive tendency map; acquiring real-time learning behavior stream data and evaluating the cognitive efficacy of the real-time learning behavior stream data to obtain real-time cognitive efficacy data; analyzing cognitive processing tendencies based on real-time cognitive efficacy data to obtain an updated cognitive tendency map; performing cognitive style evolution analysis based on the initial and updated cognitive tendency maps to obtain a cognitive style evolution trajectory; extracting learning elasticity features based on the cognitive style evolution trajectory and real-time cognitive efficacy data to obtain a learning style spectrum; and constructing a cognitive sensitivity analysis model based on the learning style spectrum.

9. The method for comprehensive educational evaluation based on a smart education AI model according to claim 1, characterized in that, Step S5 includes: This process involves identifying and updating unconventional learning patterns in a cognitive tendency map to obtain data on cognitive anomaly patterns. A cognitive sensitivity analysis model is used to intelligently combine these patterns into a set of personalized learning enhancement strategies. Based on this set of strategies and a cognitive development topography map, learning intervention schemes are simulated to obtain data predicting the effectiveness of these interventions. This data is then used to coordinate multi-party teaching resources to obtain integrated teaching resource data. Finally, adaptive strategies are applied to the integrated teaching resource data based on the cognitive transition pattern map to generate a comprehensive educational evaluation report. This report is then used to provide learning guidance.

10. A system for comprehensive educational evaluation based on a smart education AI model, used to implement the method for comprehensive educational evaluation based on a smart education AI model as described in any one of claims 1-9, characterized in that, include: Data fusion module: Collects multimodal cognitive data and learning behavior trajectories, and performs multi-source coupling of multimodal cognitive data and learning behavior trajectories through contextual intelligence integration to obtain the educational interaction information spectrum; Cognitive Topology Construction Module: Decodes the cognitive structure of the educational interaction information spectrum to obtain a cognitive graph of learning behavior, constructs the knowledge network topology based on the cognitive graph of learning behavior, and then obtains a cognitive development topography map; Cognitive Tracking Module: Collects real-time teaching interaction sequences; Cognitive fault detection is performed on the cognitive development topography to obtain a set of cognitive transition critical points, and cognitive evolution is tracked on the real-time teaching interaction sequence based on the set of cognitive transition critical points to obtain a learning development trajectory map. Model building module: Based on the learning development trajectory map, learning patterns are purified to obtain the learning style spectrum, and a cognitive sensitivity analysis model is built based on the learning style spectrum; Evaluation and guidance module: Intelligent combination of learning navigation schemes through cognitive sensitivity analysis model to obtain a set of personalized learning enhancement strategies; teaching coordination schemes are arranged according to the set of personalized learning enhancement strategies to obtain a comprehensive education evaluation report, and learning guidance is provided based on the comprehensive education evaluation report.

Images