Behavior analysis method and system based on user learning interest

Abstract

The invention relates to the technical field of data analysis, in particular to a user learning interest-based behavior analysis method and system. The method comprises the following steps: constructing a cognitive map associated with culture resources according to culture resource data to generate a culture resource cognitive map; performing user culture resource cognitive behavior analysis according to the user culture resource interaction log and the culture resource cognitive map to generate user culture resource cognitive behavior data; performing evolution characteristic analysis of user interest content increment according to the user culture resource cognitive behavior data, and generating user culture resource interest evolution characteristic data; and performing learning interest behavior prediction processing on the user culture resources based on the user culture resource interest evolution characteristic data to generate user culture resource learning interest behavior prediction data. The intelligent analysis of the learning interest behaviors of the user can be accurately realized through the information characteristics of the cultural resources and the behavior characteristics of the user.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method and system for analyzing user learning interests and behaviors. Background Technology

[0002] With the rapid growth of digital cultural resources and the widespread adoption of online learning platforms, users' learning and interaction behaviors within these systems are becoming increasingly diverse, personalized, and dynamic. Intelligent analysis techniques, such as big data analytics and machine learning, can quickly reveal users' learning interests and behaviors. For example, data collected on user browsing, clicking, and saving activities can be used as input for modeling and intelligent analysis of user learning interests. However, existing methods for analyzing user learning interests lack systematic modeling of the semantic structure and cognitive characteristics of cultural resources. This results in insufficient semantic dimensions in the analysis, difficulty in accurately reflecting users' cognitive load and learning depth, and a lack of predictive mechanisms for the dynamic evolution of user interests and potential cognitive paths. Consequently, they cannot achieve temporal modeling and evolutionary trend analysis of learning interests. Summary of the Invention

[0003] Based on this, the present invention provides a method and system for analyzing user learning interest behavior to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method based on user learning interest and behavior analysis includes the following steps: Step S1: Obtain cultural resource data from the cultural resource system; construct a cognitive map of cultural resource associations based on the cultural resource data to generate a cultural resource cognitive map; Step S2: Obtain user cultural resource interaction logs from the cultural resource system; analyze user cultural resource cognitive behavior based on the user cultural resource interaction logs and the cultural resource cognitive map, and generate user cultural resource cognitive behavior data. Step S3: Analyze the evolutionary characteristics of the incremental user interest content based on user cultural resource cognitive behavior data, and generate user cultural resource interest evolutionary characteristic data; Step S4: Analyze the potential cognitive path set of user interests based on the user cultural resource interest evolution characteristic data to generate potential cognitive path set data of user interests; perform learning interest behavior prediction processing on the potential cognitive path set data of user interests to generate learning interest behavior prediction data of user cultural resources.

[0005] Furthermore, step S1 includes the following steps: Step S11: Obtain cultural resource data from the cultural resource system; Step S12: Perform cultural resource text parsing processing on the cultural resource data to generate cultural resource text parsing data; Step S13: Perform multidimensional semantic feature analysis on the text parsing data of cultural resources to generate multidimensional semantic feature data of cultural resources; Step S14: Based on the preset attention mechanism, perform cultural resource attention cognitive characteristic analysis on the multidimensional semantic feature data of cultural resources, and generate cultural resource attention cognitive characteristic data; Step S15: Perform cluster analysis of cultural resource cognitive characteristics based on cultural resource attention cognitive characteristics data to generate cultural resource cognitive characteristics cluster data; Step S16: Design cultural resource cognitive nodes based on clustering data of cultural resource cognitive characteristics, and conduct cultural resource cognitive path analysis through cultural resource data and clustering data of cultural resource cognitive characteristics to generate cultural resource cognitive path data; Step S17: Construct a cognitive graph of cultural resources based on the cognitive nodes and cognitive paths of cultural resources, and generate a cognitive graph of cultural resources.

[0006] Furthermore, step S13 includes the following steps: Step S131: Perform the analysis of the thematic entity features of cultural resources based on the text parsing data of cultural resources, and generate thematic entity feature data of cultural resources; Step S132: Perform cultural resource entity attribute analysis on the text parsing data of cultural resources to generate cultural resource entity attribute data; Step S133: Perform semantic dependency analysis on entity attributes using cultural resource entity attribute data to generate semantic dependency data on entity attributes, and perform relation feature analysis on cultural resource attributes based on the semantic dependency data on entity attributes to generate relation feature data on cultural resource attributes. Step S134: Extract the emotionally related entities of cultural resources based on the entity attribute data of cultural resources to obtain the emotionally related entity data of cultural resources; Step S135: Based on the emotional association entity data of cultural resources, perform emotional semantic feature analysis on the text parsing data of cultural resources to generate emotional semantic feature data of cultural resources; Step S136: Perform multi-dimensional semantic feature aggregation processing on the cultural resource theme entity feature data, cultural resource attribute relationship feature data, and cultural resource emotional semantic feature data to generate multi-dimensional semantic feature data of cultural resources.

[0007] Furthermore, step S2 includes the following steps: Step S21: Obtain the user cultural resource interaction logs from the cultural resource system; Step S22: Map user cultural resource interaction logs to a cultural resource cognitive map to conduct a preliminary analysis of user cultural resource cognitive status and generate preliminary user cultural resource cognitive status data; Step S23: Perform user cultural resource interaction profile analysis based on user cultural resource interaction logs to generate user cultural resource interaction profile data; Step S24: Based on the user cultural resource interaction profile data, perform a user resource cognitive state change characteristic analysis on the preliminary user cultural resource cognitive state data to generate user resource cognitive state change characteristic data. Step S25: Perform user cultural resource cognitive load assessment processing through user cultural resource interaction profile data to generate user cultural resource cognitive load assessment data; Step S26: Analyze user cultural resource cognitive behavior based on user resource cognitive status change characteristics data and user cultural resource cognitive load assessment data to generate user cultural resource cognitive behavior data.

[0008] Furthermore, step S25 includes the following steps: Step S251: Analyze the interaction characteristics of user cultural resource types based on user cultural resource interaction profile data, and generate user cultural resource type interaction characteristic data; Step S252: Perform user cultural resource interaction specificity analysis based on user cultural resource type interaction characteristic data to generate user cultural resource interaction specificity data; Step S253: Analyze the interaction time series data based on the user cultural resource interaction profile data to generate user cultural resource interaction time series data; Step S254: Based on the user cultural resource interaction time series data, perform user cultural resource interaction specificity time series pattern analysis on the user cultural resource interaction specificity data to generate user cultural resource interaction specificity time series pattern data. Step S255: Based on the user cultural resource interaction-specific time series pattern data, perform user cultural resource cognitive load assessment processing to generate user cultural resource cognitive load assessment data.

[0009] Furthermore, step S3 includes the following steps: Step S31: Based on user cultural resource cognitive behavior data and cultural resource cognitive characteristic clustering data of cultural resource cognitive graph, perform user cultural resource interest cluster analysis to generate user cultural resource interest cluster data. Step S32: Based on the user's cultural resource interest cluster data, perform derivative association data analysis on the cultural resource cognitive map to generate interest cluster derivative association data; Step S33: Perform interest intensity assessment on the interest clusters and derived data of the interest clusters to generate interest cluster derived intensity assessment data; Step S34: Analyze the evolutionary characteristics of user interest content increments in user cultural resource interest cluster data using interest cluster derivative intensity assessment data, and generate user cultural resource interest evolutionary characteristic data.

[0010] Furthermore, step S33 includes the following steps: Step S331: Perform mutual information analysis on the derived data based on the interest cluster-derived association data to generate interest cluster-derived mutual information data; Step S332: Analyze the interest propagation characteristics of the derived data based on the interest cluster-derived related data, and generate interest cluster-derived propagation characteristic data; Step S333: Based on the mutual information data derived from interest clusters and the propagation characteristic data derived from interest clusters, perform interest intensity assessment processing on interest clusters and derived data to generate interest cluster derivation intensity assessment data.

[0011] Furthermore, step S4 includes the following steps: Step S41: Analyze the set of potential cognitive paths of user interests based on the evolutionary characteristics of user cultural resource interests, and generate the set of potential cognitive paths of user interests data; Step S42: Perform a correlation analysis on the user's interest intensity and cognitive difficulty in the set of potential cognitive paths of user interests, and generate user interest intensity-cognitive difficulty correlation data; Step S43: Design the user cultural resource interest exploration strategy space based on the correlation data of user interest intensity and cognitive difficulty, and generate user cultural resource interest exploration strategy space data; Step S44: Analyze the gradient distribution of the utility of interest exploration strategies based on the spatial data of user cultural resource interest exploration strategies, and generate gradient distribution data of the utility of interest exploration strategies. Step S45: Based on the spatial data of user cultural resource interest exploration strategies and the gradient distribution data of interest exploration strategy utility, establish the learning mapping relationship between interest exploration and utility, and generate a user cultural resource interest exploration learning model; Step S46: Use the user cultural resource interest exploration learning model to perform user cultural resource learning interest behavior prediction processing on the user interest potential cognitive path set data, and generate user cultural resource learning interest behavior prediction data.

[0012] Furthermore, step S42 includes the following steps: Step S421: Perform cultural resource cognitive information entropy analysis on the cultural resource cognitive map to generate cultural resource cognitive information entropy data; Step S422: Analyze the difficulty of users' cultural resource cognition based on user cultural resource cognitive behavior data and cultural resource cognitive information entropy data, and generate user cultural resource cognition difficulty data; Step S423: Based on the interest cluster derivative intensity assessment data and the user cultural resource cognitive difficulty data of the user interest potential cognitive path set data, perform a correlation analysis on the user interest intensity and cognitive difficulty of the user cultural resource cognitive path set data to generate user interest intensity-cognitive difficulty correlation data.

[0013] This specification also provides a user learning interest and behavior analysis system for performing the user learning interest and behavior analysis method described above. The user learning interest and behavior analysis system includes: The cultural resource analysis module is used to acquire cultural resource data from the cultural resource system; and to construct a cognitive map of cultural resource associations based on the cultural resource data, thereby generating a cultural resource cognitive map. The User Cultural Resource Cognitive Behavior Analysis Module is used to obtain user cultural resource interaction logs from the cultural resource system; and to perform user cultural resource cognitive behavior analysis based on the user cultural resource interaction logs and cultural resource cognitive maps to generate user cultural resource cognitive behavior data. The User Cultural Resource Interest Evolution Module is used to analyze the evolutionary characteristics of user interest content increments based on user cultural resource cognitive behavior data, and generate user cultural resource interest evolution characteristic data. The user learning interest and behavior prediction module is used to analyze the set of potential cognitive paths of user interests based on the evolutionary characteristics of user cultural resource interests, and generate the set of potential cognitive paths of user interests data; and to perform learning interest and behavior prediction processing on the set of potential cognitive paths of user interests data to generate learning interest and behavior prediction data of user cultural resources.

[0014] The beneficial effects of this application are as follows: This invention comprehensively acquires various cultural resource data from a cultural resource system, first performing detailed text analysis on the resource data to mine deep semantic features of cultural resources from multiple dimensions, including thematic entities, entity attributes and attribute relationships, emotionally related entities, and emotional semantics. These multi-dimensional features are then effectively aggregated to form a comprehensive characterization of the semantics of cultural resources. Based on this, attention mechanisms are used to accurately capture the attentional cognitive characteristics of cultural resources, and cluster analysis is then used to classify and integrate cultural resources with similar cognitive characteristics. Dedicated cultural resource cognitive nodes are designed based on the clustering results, and the inherent connections between cultural resources are analyzed in depth to identify reasonable cognitive paths. Finally, a complete, clearly connected, and cognitively rich cultural resource cognitive map is constructed. This not only achieves a systematic modeling of the semantic structure and cognitive characteristics of cultural resources, filling the gap of single semantic dimensions in traditional resource organization, but also provides high-quality basic data support for subsequent accurate analysis of user interaction behavior with cultural resources and the mining of user cognitive states, making the association logic of cultural resources more consistent with users' cognitive habits. By collecting various user interaction logs from the cultural resource system and combining the log data with the constructed cultural resource cognitive map, the system can quickly locate the user's current initial cognitive state, laying an initial benchmark for subsequent analysis. Based on the interaction logs, a comprehensive user interaction profile is constructed, clearly presenting the user's preferences in cultural resource selection, interaction frequency, and interaction type. Then, based on the profile, the system deeply analyzes the dynamic changes in the user's cognitive state, accurately capturing the evolution trajectory of the user's cognitive level. Simultaneously, the system extracts user interaction characteristics for different types of cultural resources from the interaction profile, uncovering specific behaviors, and combining this with interaction time-series data obtained from time-series analysis to identify the time-series change patterns of user interaction specificities, thereby scientifically assessing the user's cognitive load during the learning process. By comprehensively and accurately extracting user cultural resource cognitive behavior data from the combined changes in user cognitive state and the cognitive load assessment results, the system effectively solves the problem of quantifying user cognitive load and learning depth in the analysis. Combining user cultural resource cognitive behavior data with cognitive characteristic clustering data in the cultural resource cognitive map, the system can accurately identify the user's core interest directions in the field of cultural resources, forming structured user interest clusters and systematically integrating previously scattered user interest points. By deeply exploring the related derivative content within the cultural resource cognitive map based on this interest cluster, we can comprehensively capture users' potential extended interests and overcome the limitations of interest analysis. Analyzing the mutual information of derivative data clarifies the degree of correlation between different derivative interests and core interests; by leveraging the interest dissemination characteristics of derivative data, we can grasp the diffusion patterns of interests in the user's cognitive process. Based on this, interest intensity assessments and quantitative representations of core and derivative interests ensure the scientific rigor and accuracy of the assessment results.Based on quantified interest intensity data, this study delves into the incremental evolution patterns of user interest content. By analyzing user cultural resource interest evolution characteristics, various potential cognitive paths followed by users during cultural resource learning can be identified, forming a comprehensive set of potential cognitive paths. By analyzing the cognitive information entropy of the cultural resource cognitive map, the cognitive complexity of different cultural resources can be scientifically quantified. Combined with user cognitive behavior data, the cognitive difficulty of various resources can be accurately determined. Furthermore, a correlation analysis between cognitive difficulty and interest intensity reveals the learning adaptability of users in different interest directions. An interest exploration strategy space is designed based on the correlation between interest intensity and cognitive difficulty, covering user learning needs in different scenarios. By analyzing the utility gradient distribution of strategies, the applicable scope and expected effects of various strategies can be clarified, thereby establishing a precise learning mapping model between interest exploration and utility, ensuring that the model fully adapts to users' personalized cognitive characteristics. By using this model to predict learning interest behaviors on a set of potential cognitive paths, it is possible to accurately predict users' future learning interests and cognitive path choices. This not only provides cultural resource platforms with accurate personalized recommendation criteria and helps users plan learning paths that better suit their cognitive abilities and interests, but also improves the matching efficiency of cultural resources and the user's learning experience.

[0015] Therefore, the user learning interest behavior analysis method of this invention constructs a cognitive graph model of cultural resources by performing text parsing, semantic feature extraction, and attention-driven cognitive characteristic clustering analysis on cultural resource data. This model not only reflects the thematic, emotional, and attribute dependencies among cultural resources but also forms multi-dimensional cognitive nodes and cognitive paths, providing a high-semantic resolution cognitive semantic foundation for subsequent user interest behavior modeling. At the user behavior analysis layer, a cognitive state change characteristic analysis and cognitive load assessment mechanism are introduced. This combines the time-series behavioral characteristics, interaction-specific patterns, and cognitive response intensity of users during cultural resource interaction, reflecting the learning depth and cognitive pressure of users under different cultural resource types and complexity conditions, providing a behavioral-cognitive coupling basis for interest evolution analysis. Through fusion modeling based on the cultural resource cognitive graph and user cognitive behavior data, an interest cluster evolution analysis model is established. It extracts dynamic evolutionary features of interest intensity from the perspectives of interest derivative associations, propagation characteristics, and mutual information, capturing not only local enhancement trends of interest but also supporting the detection of cross-theme interest migration, achieving interpretable analysis of dynamic interest evolution. By combining a two-dimensional correlation analysis of interest intensity and cognitive difficulty, an interest exploration strategy space is constructed. The strategy utility gradient distribution is used to realize a mapping learning mechanism between learning utility and interest exploration, which dynamically predicts users' future learning interest behavior and provides intelligent decision-making basis for personalized recommendations, interest guidance and learning path optimization. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of a user learning interest and behavior analysis method according to the present invention. Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S4. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. Functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods. The term "and / or" as used herein includes any and all combinations of one or more of the associated items listed.

[0019] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a method for analyzing user learning interest behavior. In the embodiments of this invention, please refer to... Figure 1 The diagram shown is a flowchart illustrating the steps of a user learning interest behavior analysis method according to the present invention. The user learning interest behavior analysis method includes the following steps: Step S1: Obtain cultural resource data from the cultural resource system; construct a cognitive map of cultural resource associations based on the cultural resource data to generate a cultural resource cognitive map; In this embodiment of the invention, cultural resource data from a cultural resource system is acquired, encompassing historical document texts, subtitle texts of thematic documentary videos, explanatory texts of memorial sites, and scripts of thematic literary and artistic works. This data is then parsed using a word segmentation tool from natural language processing, loading a custom lexicon containing concepts specific to the field of red culture. Word segmentation is completed using a bidirectional maximum matching algorithm, followed by part-of-speech tagging and entity boundary recognition using a conditional random field model, generating structured text parsing results. Based on the parsing results, multi-dimensional semantic feature analysis is conducted. Thematic entity features are extracted using a word frequency-inverse document frequency algorithm combined with a pre-trained BERT model for the cultural domain, with the hidden layer dimension set to 768, outputting vector representations of entities. Entity attribute analysis employs dependency parsing tools to identify attribute relation words such as "formed from" and "related to," constructing a mapping relationship between entities and attributes. Sentimental semantic feature analysis combines a sentiment lexicon for the red culture domain (containing positive words such as "exhilarating" and "lofty," and specific sentiment words such as "tragic"), calculating sentiment characteristics by weighted summation of positive word counts and specific sentiment word counts divided by the total number of words. Three types of features are concatenated to form a multi-dimensional semantic feature vector, which is then input into a Transformer-based attention mechanism module. This module has a hidden layer dimension of 512 and 8 attention heads. By calculating the attention weights between features, easily noticed cognitive characteristics in cultural resources are extracted. Based on the attention weights, a density clustering algorithm is used, with weighted Euclidean distance as the similarity measure and a cluster radius of 0.3, to group resources with similar cognitive characteristics into one class, generating clustering results. Cognitive nodes are designed based on the central feature vector of each cluster. Each node contains a topic entity label, a set of core attributes, and a sentiment tendency value. Co-occurrence relationships between resources are analyzed using an association rule mining algorithm, with a support threshold of 0.1 and a confidence threshold of 0.5, to filter out significantly associated resource pairs as cognitive paths. Finally, using a graph database, cognitive nodes are used as vertices of the graph, and cognitive paths are used as edges with association strength (confidence value), constructing a cultural resource cognitive graph.

[0020] Step S2: Obtain user cultural resource interaction logs from the cultural resource system; analyze user cultural resource cognitive behavior based on the user cultural resource interaction logs and the cultural resource cognitive map, and generate user cultural resource cognitive behavior data. In this embodiment of the invention, user interaction logs from a cultural resource system are acquired, including user click records on cultural resources, page dwell time (accurate to the second), collection operation markers, note text content, resource jump paths, etc. Resource IDs in the logs are matched with node IDs in a cultural resource cognitive graph. The frequency of user interaction nodes over 7 days is statistically analyzed. Nodes with a frequency exceeding 30% of the total interaction volume are designated as core cognitive nodes, forming a preliminary cognitive state. A user interaction profile is constructed based on the logs, including the set of interaction nodes, the interaction frequency percentage of each node, the average dwell time distribution, and the percentage of different types of operations (clicks / collections / notes). The intersection-union ratio (IUU) of the core cognitive node sets for two consecutive weeks is calculated. If the IUU is below 0.5, a change in cognitive state is determined, and the type and interaction frequency of newly added nodes are recorded, generating state change characteristic data. When conducting cognitive load assessment, the frequency of user interactions and average dwell time with different types of cultural resources (such as documentary texts, documentary videos, and introductions to memorial sites) are first extracted to construct a two-dimensional feature matrix. The interaction-specific data are obtained by calculating the deviation of each feature value from the system average feature value (using standardized Z-scores, with an absolute deviation greater than 1.5 considered specific). The interaction logs are divided into time-series segments at the hourly granularity, and a Hidden Markov Model (HMM) is used to identify time-series patterns in the specific data. The model sets five hidden states, and the state transition probabilities are calculated using a forward-backward algorithm to generate time-series patterns. Based on the duration and transition frequency of specific states in the patterns, a load score (the proportion of specific state duration multiplied by the transition frequency) is calculated; a higher score indicates a greater cognitive load. Finally, user cognitive state change characteristic data (including change time points and new node features) and cognitive load assessment data (load score and high-load periods) are integrated to form user cultural resource cognitive behavior data.

[0021] Step S3: Analyze the evolutionary characteristics of the incremental user interest content based on user cultural resource cognitive behavior data, and generate user cultural resource interest evolutionary characteristic data; In this embodiment of the invention, based on the interaction frequency, dwell time, and clustering data of the cultural resource cognitive graph in user cultural resource cognitive behavior data, the K-means clustering algorithm is used to analyze interest clusters. The cosine similarity between interaction frequency and cluster centers is used as a distance metric. The optimal number of clusters is determined by the silhouette coefficient (the number of clusters with a silhouette coefficient higher than 0.7 is the final value). Resources with high-frequency user interaction and similar characteristics are clustered into interest clusters. Each cluster contains a core resource ID and an average interaction intensity (frequency multiplied by duration). Based on the core resource nodes in the interest clusters, their directly adjacent nodes (first-degree association) and adjacent nodes connected through one intermediate node (second-degree association) are traversed in the cultural resource cognitive graph. The features of these nodes and their association strength with the core nodes are extracted to generate interest cluster derived association data. To assess the interest intensity of derived related data, the mutual information value between derived nodes and core cluster nodes is first calculated (using the ratio of joint probability to marginal probability). Then, the PageRank algorithm is used to analyze the influence of derived nodes in interest propagation. The algorithm iterates 100 times with a damping coefficient of 0.85 to obtain propagation characteristic values. The mutual information value and propagation characteristic value are weighted and summed in a 6:4 ratio to obtain the interest intensity assessment result. Using a weekly time window, the rate of change of core strength and derived strength of each interest cluster is calculated (current window strength minus previous window strength, then divided by the previous window strength). Derivative nodes with positive intensity growth and their growth rate are captured. The characteristic association between newly added derived nodes and existing clusters is analyzed to generate user cultural resource interest evolution characteristic data containing the expansion direction of interest clusters, intensity change trends, and characteristics of newly added derived nodes.

[0022] Step S4: Analyze the potential cognitive path set of user interests based on the user cultural resource interest evolution characteristic data to generate potential cognitive path set data of user interests; perform learning interest behavior prediction processing on the potential cognitive path set data of user interests to generate learning interest behavior prediction data of user cultural resources.

[0023] In this embodiment of the invention, based on the user's cultural resource interest evolution characteristic data, node sequences with continuously increasing interest intensity within the last three time windows are extracted. Combined with cognitive paths in the cultural resource cognitive graph, paths with a correlation strength higher than 0.6 between nodes are selected to form a set of potential cognitive paths for user interests. Each path includes a node sequence, total path length (number of nodes), and average correlation strength. For each node in the cultural resource cognitive graph, the Shannon entropy formula is used to calculate cognitive information entropy (a measure of uncertainty based on node attributes; the more attribute categories and the more uniform the distribution, the higher the entropy value). Combined with the average dwell time of the node in the user's cognitive behavior data (the longer the dwell time, the greater the cognitive difficulty), the product of the information entropy value and the dwell time is used as the user's cognitive difficulty value for that node. The Pearson correlation coefficient is calculated between the interest intensity assessment value and the cognitive difficulty value of each node in the potential cognitive path to obtain the interest intensity-cognitive difficulty correlation data (correlation coefficient and significance level) for each path. Based on associated data, the interest exploration strategy space is divided into four quadrants according to interest intensity (high / low) and cognitive difficulty (high / low). Each quadrant corresponds to a specific strategy (e.g., the strategy for the high interest-low difficulty quadrant is "deep learning," and the strategy for the high interest-high difficulty quadrant is "step-by-step learning"). The prediction accuracy of each strategy in historical data (the percentage of correctly predicted user-selected paths) is calculated. The rate of change of accuracy is calculated using the sliding window method (window size of 50 historical data points) to obtain the gradient distribution of interest exploration strategy utility. Based on the strategy space data and utility gradient data, a random forest model is constructed. The model is set with 200 decision trees and a maximum depth of 10. The parameters are optimized through 5-fold cross-validation. The node features of the path, interest intensity, and cognitive difficulty are used as input features, and strategy utility is used as the output label to generate a user cultural resource interest exploration learning model. The set of potential cognitive paths of user interests is input into the model, and the predicted occurrence probability of each path is output. Paths with a probability higher than 0.5 are included in the final user cultural resource learning interest behavior prediction data, which includes path sequence, prediction probability, and recommendation priority (arranged in descending order of probability).

[0024] Furthermore, step S1 includes the following steps: Step S11: Obtain cultural resource data from the cultural resource system; In this embodiment of the invention, cultural resource data from a cultural resource system is acquired, including, for example, documentary texts from the revolutionary historical period, subtitles of red-themed video materials, physical descriptions of memorial sites, and textual materials from red-themed literary and artistic creations. The data covers different historical periods and different forms of expression, ensuring multi-dimensional content encompassing ideology, historical events, figures, and spiritual connotations. During the acquisition process, the data undergoes integrity verification, removing duplicate content and incomplete text to form a structured raw data set, providing a foundation for subsequent text analysis and feature extraction.

[0025] Step S12: Perform cultural resource text parsing processing on the cultural resource data to generate cultural resource text parsing data; In this embodiment of the invention, text parsing processing of cultural resource data is performed using a word segmentation tool from the field of natural language processing. This tool loads a custom vocabulary containing culturally specific concepts (such as names of specific historical periods, organizational names, and theoretical terms), and the vocabulary is updated periodically to cover newly added concepts. The text is segmented using a bidirectional maximum matching algorithm, scanning the text first from left to right and then from right to left, selecting the segmentation result with the most matching words. After segmentation, a conditional random field model is used for part-of-speech tagging. The model is trained and optimized using a cultural corpus, and the tagging categories include nouns, verbs, adjectives, and proper nouns. Simultaneously, an entity boundary recognition algorithm is used to determine the start and end positions of entities such as historical events, figures, and organizations based on the part-of-speech tagging results and the semantic association with the context. The final generated cultural resource text parsing data includes the word segmentation sequence, the part-of-speech tag for each word, the entity type, and the entity's position index in the text, forming a structured parsing result.

[0026] Step S13: Perform multidimensional semantic feature analysis on the text parsing data of cultural resources to generate multidimensional semantic feature data of cultural resources; In this embodiment of the invention, multi-dimensional semantic feature analysis is performed based on the parsed text data of cultural resources. Thematic entity feature analysis employs a word frequency-inverse document frequency algorithm, statistically analyzing the frequency of each entity's occurrence in a single text and its distribution frequency across all texts, and calculating feature weights. Simultaneously, a pre-trained BERT model for the red culture domain is combined. This model, fine-tuned with 100,000 red culture texts and a hidden layer dimension of 768, inputs entity text fragments into the model and outputs a 768-dimensional vector as the deep semantic representation of the entity. The two are weighted and fused to form thematic entity feature data. Entity attribute analysis is achieved through dependency parsing tools. These tools identify attribute relation words such as "formed in," "belongs to," and "manifests as" in the text, constructing mapping relationships between entities and attributes, such as "event-occurrence time" and "person-affiliated organization," and determining the attribute association strength based on the frequency of relation words, generating entity attribute data. Entity attribute semantic dependency analysis calculates the co-occurrence probability between attributes, identifying dependency relationships such as "time-place" and "cause-effect," forming attribute relationship feature data. Sentiment-related entity extraction is based on the co-occurrence distance between entities and sentiment words (limited to 5 words), filtering entities closely associated with sentiment words. Sentiment semantic feature analysis is achieved through sentiment tendency value calculation. This involves counting the number of positive sentiment words and specific sentiment words (such as words expressing tragedy or passion) in the text, weighted 8:2, summing them, and then dividing by the total number of words to obtain the sentiment tendency value. Finally, the feature vectors of topic entities, attribute relationship feature vectors, and sentiment semantic feature values ​​are concatenated to form multidimensional semantic feature data of cultural resources.

[0027] Step S14: Based on the preset attention mechanism, perform cultural resource attention cognitive characteristic analysis on the multidimensional semantic feature data of cultural resources, and generate cultural resource attention cognitive characteristic data; In this embodiment of the invention, attentional cognitive characteristics of multidimensional semantic feature data of cultural resources are analyzed based on a preset attention mechanism. A multi-head self-attention module in the Transformer architecture is used, with a hidden layer dimension of 512 and divided into 8 attention heads, each independently processing a 64-dimensional feature subspace. The multidimensional semantic feature data (1024-dimensional vector) is transformed into query vector, key vector, and value vector through linear projection, all three with a dimension of 512. The dot product of the query vector and key vector is calculated, the result is scaled by dividing by the square root of the feature dimension, and then normalized using the softmax function to obtain the attention weight matrix. Each element in the matrix represents the attention level of the corresponding feature, with weight values ​​limited to the range of 0-1 and a row sum of 1. The weight matrix is ​​multiplied by the value vector to obtain the weighted semantic representation of each feature. The outputs of the 8 attention heads are concatenated and subjected to linear transformation to generate the final attention feature vector. During the analysis, features with weight values ​​exceeding 0.7 are extracted as high-attention cognitive characteristics. These features include core themes, sentiment keywords, and key attribute descriptions in cultural resources. The generated cultural resource attention and cognitive characteristics data includes feature indexes, corresponding attention weight values, high attention markers, and feature semantic descriptions, fully reflecting the cognitive points in cultural resources that are easily noticed by users and their importance.

[0028] Step S15: Perform cluster analysis of cultural resource cognitive characteristics based on cultural resource attention cognitive characteristics data to generate cultural resource cognitive characteristics cluster data; In this embodiment of the invention, cognitive characteristic clustering analysis is performed based on cultural resource attention and cognitive characteristic data. A density clustering algorithm is used, with the Euclidean distance between attention feature vectors as the similarity criterion. The distance is calculated as the square root of the sum of the squares of the differences in each dimension of the vectors; a smaller distance indicates that the cognitive characteristics of the two resources are more similar. The algorithm sets the cluster radius to 0.3 (determined based on 1 / 3 of the standard deviation of the feature vectors to ensure that the differences in characteristics of similar resources are within a controllable range), and the minimum number of samples is 5 (to avoid meaningless clustering due to an insufficient sample size). The clustering process first traverses all unlabeled resources, randomly selects a resource as the initial core point, and searches for all resources within a radius of 0.3 centered on the core point to form a temporary cluster. If the number of resources in the temporary cluster reaches or exceeds 5, it is determined as a valid cluster, and all resources within the cluster are labeled; otherwise, the core point is labeled as a noise point. This process is repeated until all resources are labeled as members of a certain cluster or noise points. After clustering is completed, the central feature vector of each cluster is calculated, and the theme overlap and high attention feature distribution of resources within the cluster are statistically analyzed. This generates cultural resource cognitive characteristic clustering data containing a unique cluster identifier, central feature vector, total number of resources within the cluster, and core cognitive characteristics (the top 5 high-frequency and high-attention features), thus realizing the systematic classification of cultural resources with similar cognitive characteristics.

[0029] Step S16: Design cultural resource cognitive nodes based on clustering data of cultural resource cognitive characteristics, and conduct cultural resource cognitive path analysis through cultural resource data and clustering data of cultural resource cognitive characteristics to generate cultural resource cognitive path data; In this embodiment of the invention, cognitive nodes are designed based on clustered data of cultural resource cognitive characteristics. Each node is based on the central feature vector of the corresponding cluster and includes topic entity labels, core attribute sets, sentiment values, and cognitive characteristic labels. Cognitive path analysis is performed using cultural resource data and clustered data. An association rule mining method is used, with clusters as the analysis unit, to count the co-occurrence frequency of different clusters in the user's historical interaction sequence (i.e., the number of times a user continuously accesses resources in different clusters). A support threshold of 0.1 (the proportion of the co-occurrence frequency of a certain cluster combination to the total number of interaction sequences) and a confidence threshold of 0.5 (the probability that if a user visits cluster A, they will subsequently visit cluster B) are set, and association rules that simultaneously satisfy both thresholds are selected. Each rule corresponds to a cognitive path, with the starting point of the path being the cluster preceding the rule and the ending point being the cluster following the rule. The path strength is the confidence value, and the path directionality is marked (e.g., A→B and B→A are two independent paths). The generated cognitive path data for cultural resources includes starting node identifiers, ending node identifiers, path strength, path direction, and co-occurrence frequency, clearly reflecting the cognitive relationship logic between cultural resources.

[0030] Step S17: Construct a cognitive graph of cultural resources based on the cognitive nodes and cognitive paths of cultural resources, and generate a cognitive graph of cultural resources.

[0031] In this embodiment of the invention, a cognitive graph of cultural resource associations is constructed based on cognitive node and cognitive path data. A graph database is used as the carrier, supporting vertex attribute storage, weighted directed edge definition, and hierarchical relationship construction. Each cognitive node is mapped to a vertex in the graph. Vertex attribute fields include a unique node identifier, topic entity label, core attribute set, sentiment tendency value, cognitive characteristic label, and cluster identifier. Attribute values ​​are stored in a structured format (e.g., the core attribute set is recorded as key-value pairs). Cognitive path data is mapped to directed edges in the graph. Edge attribute fields include starting vertex identifier, ending vertex identifier, path strength (confidence value), path direction identifier, and co-occurrence frequency. During the construction process, node deduplication is performed: the cosine similarity of the central feature vectors of different nodes is calculated. If the similarity is greater than 0.9 (indicating high overlap of cognitive characteristics), they are merged into the same vertex, retaining node information with more complete attributes. Path redundancy is eliminated: for multiple paths with the same starting and ending points, if the difference in path strength is less than 0.1, only the path with the highest strength is retained. Simultaneously, the degree value (total number of associated edges) of each vertex is calculated. Vertices with a degree value higher than 1.5 times the average degree value of all vertices are marked as core nodes, serving as key connection hubs in the graph. The final generated cultural resource cognitive graph includes a set of vertices (including attributes), a set of edges (including attributes), and node hierarchical relationships (parent and child nodes are determined based on attribute relationships such as "whole-part" and "genus-species"), forming a structured cultural resource cognitive association network that intuitively presents the cognitive logic and association strength between resources.

[0032] Furthermore, step S13 includes the following steps: Step S131: Perform the analysis of the thematic entity features of cultural resources based on the text parsing data of cultural resources, and generate thematic entity feature data of cultural resources; In this embodiment of the invention, topic entity feature analysis is performed based on the parsed text data of cultural resources, employing a combination of the term frequency-inverse document frequency (IF-IVF) algorithm and a pre-trained language model. Based on the entity annotation results in the parsed data, the frequency of each entity in a single text (term frequency) and the proportion of texts containing that entity to the total text volume (document frequency) are calculated. The basic feature weight of the entity is calculated by multiplying the term frequency by the IVF (the logarithm of the ratio of the total number of texts to the number of texts containing that entity). Simultaneously, a BERT model fine-tuned with 100,000 cultural domain texts is used, with the hidden layer dimension set to 768. The text fragment corresponding to each entity (including the context of 10 words before and after the entity) is input into the model, and the average vector of the last hidden state is extracted as the deep semantic representation of the entity. The basic feature weights and the deep semantic representation are weighted and fused in a 3:7 ratio to generate a 768-dimensional topic entity feature vector. The final cultural resource topic entity feature data includes a unique entity identifier, feature vector, feature weight value, and a core contextual description of the entity in the text, comprehensively depicting the semantic features of core topic entities in cultural resources.

[0033] Step S132: Perform cultural resource entity attribute analysis on the text parsing data of cultural resources to generate cultural resource entity attribute data; In this embodiment of the invention, entity attribute analysis is performed on the parsed text data of cultural resources using a dependency parsing tool combined with an attribute relation word list for the field of Red Culture. The tool analyzes the syntactic structure of the text to identify pre-defined attribute relation words such as "formed in," "belongs to," "embodied in," and "occurs in." These relation words are categorized by attribute type in the word list (e.g., time-based, organizational, connotation-based, and regional). For each identified entity, the object component connected by the relation word is traced to determine the entity's attribute value. The co-occurrence frequency of the same entity and the same attribute value is statistically analyzed; attribute pairs with a frequency higher than 5 are identified as stable attribute associations, and the frequency is used as the attribute association strength (normalized to the 0-1 range). The generated cultural resource entity attribute data includes entity identifier, attribute type, attribute value, association strength, and the original expression of the attribute in the text, clearly presenting the core attribute characteristics of Red Culture entities.

[0034] Step S133: Perform semantic dependency analysis on entity attributes using cultural resource entity attribute data to generate semantic dependency data on entity attributes, and perform relation feature analysis on cultural resource attributes based on the semantic dependency data on entity attributes to generate relation feature data on cultural resource attributes. In this embodiment of the invention, semantic dependency analysis of entity attributes is performed using cultural resource entity attribute data. The co-occurrence probability between different attributes is calculated. Co-occurrence refers to the simultaneous appearance of two attributes in the attribute set of the same entity; the probability is the ratio of the number of co-occurrences to the total number of attribute combinations. A co-occurrence probability threshold of 0.3 is set to filter out attribute pairs with high probability of co-occurrence. Combined with cultural domain knowledge, semantic dependencies such as "time-region," "cause-effect," and "whole-part" are identified. For each dependency, a conditional probability (the probability that attribute B exists if attribute A exists) is calculated as the relationship strength, and the strength value is normalized to the 0-1 range. Attribute relationship feature analysis is performed based on the dependency data, mapping each dependency to a feature dimension, with the dimension value representing the relationship strength, generating a 200-dimensional attribute relationship feature vector (covering all high-frequency dependency types). The final entity attribute semantic dependency data includes dependency pairs, co-occurrence probabilities, and conditional probabilities; the cultural resource attribute relationship feature data includes feature vectors and dependency descriptions corresponding to each dimension, comprehensively reflecting the semantic association patterns between attributes.

[0035] Step S134: Extract the emotionally related entities of cultural resources based on the entity attribute data of cultural resources to obtain the emotionally related entity data of cultural resources; In this embodiment of the invention, emotionally related entities are extracted based on cultural resource entity attribute data. Based on a pre-defined cultural sentiment lexicon, these are categorized into positive sentiment words and specific sentiment words according to their emotional tendency. The co-occurrence of entities and sentiment words in the entity attribute data is analyzed, with co-occurrence limited to the same text sentence and the interval between the entity and the sentiment word not exceeding 5 words. The co-occurrence frequency of each entity and sentiment word is counted, and the ratio of the co-occurrence frequency to the total occurrence frequency of the entity (emotional relevance) is calculated. A relevance threshold of 0.2 is set, and entities with a relevance higher than the threshold are selected. For these entities, their associated sentiment word type (positive / specific) and the highest frequency associated sentiment word are labeled. The generated cultural resource emotionally related entity data includes entity identifiers, associated sentiment word types, high-frequency sentiment words, and emotional relevance, accurately locating cultural entities closely related to emotional expression.

[0036] Step S135: Based on the emotional association entity data of cultural resources, perform emotional semantic feature analysis on the text parsing data of cultural resources to generate emotional semantic feature data of cultural resources; In this embodiment of the invention, emotional semantic feature analysis is performed on text parsing data based on cultural resource emotional association entity data. First, emotional words co-occurring with emotional association entities are extracted from the text, distinguishing between positive emotional words and specific emotional words. The total number of positive emotional words (denoted as N1) and the total number of specific emotional words (denoted as N2) in a single text are counted. Weights are assigned according to cultural emotional features, with a weight of 0.8 for positive emotional words and a weight of 0.2 for specific emotional words, calculating a weighted emotional value: (0.8 × N1 + 0.2 × N2). Dividing the weighted emotional value by the total number of words in the text yields the emotional tendency value; a higher value indicates a stronger emotional expression in the text. Simultaneously, the distribution density of emotional words in the text is statistically analyzed as an auxiliary feature. The generated cultural resource emotional semantic feature data includes emotional tendency value, the proportion of positive emotional words, the proportion of specific emotional words, and the distribution density of emotional words, quantitatively characterizing the emotional semantic features of cultural resource texts.

[0037] Step S136: Perform multi-dimensional semantic feature aggregation processing on the cultural resource theme entity feature data, cultural resource attribute relationship feature data, and cultural resource emotional semantic feature data to generate multi-dimensional semantic feature data of cultural resources.

[0038] In this embodiment of the invention, multi-dimensional semantic feature aggregation processing is performed on cultural resource theme entity feature data, attribute relationship feature data, and sentiment semantic feature data, using a combination of feature concatenation and standardization. The theme entity feature data is a 768-dimensional vector, the attribute relationship feature data is a 200-dimensional vector, and the sentiment semantic feature data includes four quantitative indicators (sentiment tendency value, proportion of positive sentiment words, etc.), which are converted into 4-dimensional vectors. First, the three types of feature vectors are standardized by subtracting the mean of each dimension and dividing by the standard deviation, so that all feature dimensions are of the same magnitude (mean is 0, standard deviation is 1). Then, the theme entity feature vector, attribute relationship feature vector, and sentiment semantic feature vector are concatenated in the order of theme entity feature vector, attribute relationship feature vector, and sentiment semantic feature vector to form an initial aggregated vector of 768 + 200 + 4 = 972 dimensions. Principal component analysis is used to reduce the dimensionality of the initial vector, retaining the principal components with a cumulative variance contribution rate of 95%, ultimately generating 1024-dimensional multi-dimensional semantic feature data of cultural resources.

[0039] Furthermore, step S2 includes the following steps: Step S21: Obtain the user cultural resource interaction logs from the cultural resource system; In this embodiment of the invention, user interaction logs from a cultural resource system are acquired. The logs include user click records (including unique identifiers for clicked resources), page dwell time (the time difference between resource loading completion and leaving the page), collection operation markers, note text content, and resource navigation paths. Log data is stored categorized by user unique identifier, with each record containing a user identifier, operation type, resource identifier, timestamp, and related operation information. The acquired logs are cleaned to remove invalid records: click records with a dwell time of less than 3 seconds, duplicate records with the same timestamp, and records with empty notes. The cleaned logs form a structured set of user interaction logs, providing foundational data for subsequent cognitive state analysis and profile construction.

[0040] Step S22: Map user cultural resource interaction logs to a cultural resource cognitive map to conduct a preliminary analysis of user cultural resource cognitive status and generate preliminary user cultural resource cognitive status data; In this embodiment of the invention, user cultural resource interaction logs are mapped to a cultural resource cognitive graph. A correspondence needs to be established between the unique resource identifier recorded in the logs and the resource association identifier of the cognitive graph nodes to ensure that each interaction record can be accurately located to a specific node in the graph. The graph nodes involved in all user interactions over seven consecutive days are statistically analyzed, and the interaction frequency of each node is cumulatively calculated. Clicks, favorites, and notes are all counted as one valid interaction; repeated clicks on the same node within 30 minutes are counted only as one interaction to avoid duplicate counting. The top five nodes with the highest cumulative frequency are selected as core interaction candidate nodes. The proportion of each candidate node's interaction frequency to the total interaction frequency is calculated; nodes with a proportion exceeding 30% are identified as core cognitive nodes, and the remaining candidate nodes and other interaction nodes are considered associated cognitive nodes. For core cognitive nodes, their interaction intensity is calculated using the formula: node interaction frequency multiplied by the node's average dwell time (total dwell time divided by interaction frequency). The intensity value is rounded to the nearest integer to quantify the interaction depth. The generated preliminary user cultural resource cognitive status data includes a core cognitive node set, with each node accompanied by its identifier and cognitive characteristic label in the cognitive graph; a set of related cognitive nodes, along with their interaction frequency; the interaction intensity value of the core nodes; and the total interaction frequency over 7 days. This data clearly presents the user's recent core focus, related content, and interaction depth in cultural resource learning, laying the foundation for subsequent analysis of cognitive status changes.

[0041] Step S23: Perform user cultural resource interaction profile analysis based on user cultural resource interaction logs to generate user cultural resource interaction profile data; In this embodiment of the invention, interaction profiling analysis is performed based on user cultural resource interaction logs. This requires extracting multi-dimensional indicators from the logs and integrating them in a structured manner. The interaction node set includes the graph node identifiers of all user interactions within 7 days, and the resource type corresponding to each node. Resource types are divided into three categories: document text, documentary video, and venue introduction. Each node belongs to only one type. The interaction frequency percentage is calculated by dividing the interaction frequency of a single node by the total interaction frequency. The top 10 nodes with the highest percentages are selected and sorted in descending order. The remaining nodes are merged into the "Other" category, and their combined percentage is calculated. The average dwell time distribution is statistically analyzed separately for each resource type. The total dwell time for each type is divided by the interaction frequency to obtain the average dwell time, accurate to the second. The dwell time is also divided into three intervals: 0-60 seconds, 61-180 seconds, and over 181 seconds. The proportion of interactions within each interval to the total number of interactions for that type is calculated for each interval. The percentage of each operation type was calculated as follows: clicks, favorites, and notes were counted as a percentage of the total number of interactions. For note operations, the average note length was calculated by dividing the total number of characters in all notes by the number of notes, rounded to the nearest integer. The generated user cultural resource interaction profile data is presented in a structured format, including the set of interaction nodes and the distribution of resource types, the percentage of interaction frequency of the top 10 nodes and the percentage of the "other" category, the average dwell time and interval percentage of each type of resource, the percentage of the three types of operations, and the average note length. It also includes the data update time (accurate to the hour). This data comprehensively depicts user preferences in cultural resource interactions, including the specific content they focus on, their dwell time habits for different types of resources, and their operational behavior tendencies, providing detailed evidence for subsequent cognitive state changes and workload assessment.

[0042] Step S24: Based on the user cultural resource interaction profile data, perform a user resource cognitive state change characteristic analysis on the preliminary user cultural resource cognitive state data to generate user resource cognitive state change characteristic data. In this embodiment of the invention, the cognitive state change characteristics analysis is performed on the preliminary cognitive state data based on user cultural resource interaction profile data. Two consecutive 7-day interaction cycles are selected as the analysis unit. The core cognitive node sets of the two cycles are extracted, and the intersection-union ratio (IUR) of the two sets is calculated. The formula is the number of nodes commonly included in the two sets divided by the total number of nodes in both sets (excluding duplicate nodes). An IUR below 0.5 indicates a significant change in cognitive state, while an IUR above or equal to 0.5 indicates a stable state. The time point of change, i.e., the start time of cycle two, is recorded, accurate to the day. The newly added core nodes in cycle two compared to cycle one are statistically analyzed; these are nodes that were not included in the core set in cycle one but entered the core set in cycle two. The feature difference between these newly added nodes and the core nodes in cycle one is calculated using the cosine similarity of the node feature vectors. Nodes with a similarity below 0.6 are marked as highly different, and those above or equal to 0.6 are marked as lowly different. Simultaneously, the rate of change of interaction intensity for each core node within the two cycles is calculated. Nodes with a rate of change exceeding 50% are marked as nodes with significant changes in intensity, and the magnitude of these changes is recorded. The generated user resource cognitive state change characteristic data includes state change judgment results, the time point of the change, the set of newly added core nodes and the feature difference label of each node, the set of nodes with significant changes in intensity and their respective change rates and magnitudes. Through this data, the migration trajectory of user cognitive focus can be accurately captured, including the time of change, the degree of difference between newly added content and existing content, and the change in the depth of interaction with core content, providing support for understanding the dynamic evolution of user cognitive behavior.

[0043] Step S25: Perform user cultural resource cognitive load assessment processing through user cultural resource interaction profile data to generate user cultural resource cognitive load assessment data; In this embodiment of the invention, cognitive load assessment is performed using user cultural resource interaction profile data, requiring the integration of multi-dimensional data in stages. First, interaction features of different resource types are extracted from the interaction profile, including the interaction frequency ratio, average dwell time, and note-taking ratio for three categories: document text, documentary video, and venue introduction, forming a basic feature matrix. Using the mean of the same resource type interaction features for all users in the system as the benchmark, the standardized Z-score for each user feature value is calculated. The formula is (user feature value minus system mean) divided by the system standard deviation, which is calculated based on the dispersion of that feature value across all users. A Z-score absolute value greater than 1.5 is set as an anomaly threshold, marking user anomalies, such as a frequency ratio Z-score exceeding 1.5 indicating high-frequency anomaly, and a dwell time Z-score exceeding 1.5 indicating long-term dwell anomaly. The interaction log is divided into time-series segments at an hourly granularity. Each segment records the total number of interactions, the number of resource types involved, and the number of times anomalies occur within that hour. A Hidden Markov Model (HMM) was used for pattern recognition on time-series data. The model had five hidden states representing cognitive load: initial contact, long-term low-frequency contact, short-term low-frequency contact, short-term high-frequency contact, and long-term high-frequency contact. Observations were the number of specific features in each hourly segment. The Baum-Welch algorithm was used to train the model, initializing the state transition probabilities to a uniform distribution and setting the observation probabilities based on a normal distribution of specific feature counts. The model was iteratively trained 50 times until the likelihood function value changed to less than 1e-6, reaching convergence. The Viterbi algorithm was used to decode the load state of each hourly segment. The proportion of the total duration of the load state to the total analysis time was calculated as the load percentage, and the number of state transitions per hour was calculated as the transition frequency. The cognitive load score was obtained by multiplying the load percentage by the transition frequency and then by 5, with a range limited to 0-5. The most frequently occurring specific feature types within these time periods were recorded. The generated user cultural resource cognitive load assessment data includes cognitive load score, load percentage, conversion frequency, load period list, and main unique feature types, comprehensively quantifying the cognitive pressure level, distribution pattern, and related interaction characteristics of users in cultural resource learning.

[0044] Step S26: Analyze user cultural resource cognitive behavior based on user resource cognitive status change characteristics data and user cultural resource cognitive load assessment data to generate user cultural resource cognitive behavior data.

[0045] In this embodiment of the invention, cognitive behavior analysis is performed based on user resource cognitive state change characteristics data and cultural resource cognitive load assessment data, requiring the establishment of a multi-dimensional correlation analysis framework. The time points of state changes are aligned with high-load periods, and the average cognitive load score within the 24 hours prior to the change is calculated. If the score exceeds 3, the change is marked as a "high-load driven change," indicating that the user's cognitive focus shift may be triggered by high cognitive pressure; if the score is less than or equal to 3, it is marked as a "proactive exploration change," indicating that the change is more likely to stem from the user's autonomous interest expansion. The correlation between the cognitive characteristics of newly added core nodes and the types of resources interacted with during high-load periods is analyzed. This is calculated by dividing the number of overlapping attributes by the total number of attributes. New nodes with a correlation score exceeding 0.6 are marked as "load-related new nodes," indicating a close connection between them and the learning content under high-load conditions. For periods of stable cognitive states, the Pearson correlation coefficient between the average cognitive load score and the interaction intensity of core nodes within this period is calculated. A coefficient with an absolute value exceeding 0.5 is considered a strong correlation, and the direction of the correlation (positive or negative) is recorded. A positive correlation indicates that the higher the load, the greater the interaction intensity, while a negative correlation indicates the opposite. The generated user cultural resource cognitive behavior data includes behavior type labels (high-load driven or proactive exploration), alignment results between change time points and high-load periods, attribute association details of newly added nodes associated with load, and correlation data between load and interaction intensity during stable periods. Through this data, the dynamic changes in user cognitive states and cognitive load characteristics are comprehensively integrated to accurately depict the deep behavioral patterns of users in cultural resource learning, including the driving factors of cognitive focus shifts, the correlation between new content and existing learning pressure, and the relationship between load and engagement during stable learning stages, providing core evidence for subsequent interest evolution analysis.

[0046] Furthermore, step S25 includes the following steps: Step S251: Analyze the interaction characteristics of user cultural resource types based on user cultural resource interaction profile data, and generate user cultural resource type interaction characteristic data; In this embodiment of the invention, user cultural resource interaction profile data is used to analyze the interaction characteristics of resource types. For example, resource types are explicitly divided into three categories: document text, documentary video, and venue introduction. The interaction frequency percentage for each type is calculated by dividing the total number of interactions for a single resource type by the total number of interactions for all resource types. The average dwell time for each type is calculated by dividing the total dwell time for a single resource type by the number of interactions for that type, accurate to the second. The percentage of different operations within each type is also calculated: the number of clicks, favorites, and notes are divided by the total number of interactions for that type. For note operations, the average number of characters per note is additionally calculated by dividing the total number of characters in all notes of that type by the number of notes. These indicators are systematically organized by resource type to form structured data containing type name, frequency percentage, average dwell time, click percentage, favorite percentage, note percentage, and average note character count. This data comprehensively presents users' interaction preferences for different types of cultural resources, including attention frequency, dwell habits, and operational tendencies, providing multi-dimensional basic characteristics for subsequent specific analysis.

[0047] Step S252: Perform user cultural resource interaction specificity analysis based on user cultural resource type interaction characteristic data to generate user cultural resource interaction specificity data; In this embodiment of the invention, interaction specificity analysis is performed based on user cultural resource type interaction characteristic data. The average interaction characteristic of all users of the same resource type in the system is used as the benchmark value, which includes average frequency percentage, average dwell time, and average operation percentage. A standardized Z-score is calculated for each user's characteristic value. The calculation method is to subtract the system mean from the user's characteristic value and then divide by the system standard deviation. The system standard deviation is calculated based on the dispersion of the characteristic value across all users. A Z-score absolute value greater than 1.5 is set as the specificity threshold. When the frequency percentage Z-score of a user for a certain type of resource exceeds 1.5, it is marked as high-frequency specificity; when it is lower than -1.5, it is marked as low-frequency specificity. A dwell time Z-score exceeding 1.5 is marked as long-stay specificity, and lower than -1.5 is marked as short-stay specificity. Among the operation percentages, a note percentage Z-score exceeding 1.5 is marked as high note participation specificity. The generated interaction-specific data includes resource type, specific feature name, Z-score value, and specific direction, clearly identifying interaction features that show significant differences between the user and the system average level.

[0048] Step S253: Analyze the interaction time series data based on the user cultural resource interaction profile data to generate user cultural resource interaction time series data; In this embodiment of the invention, interaction time-series data is parsed based on user cultural resource interaction profile data, and timestamps of all interaction records are extracted from the profile data. For interval segments, the total number of interactions within that time period is counted, i.e., the total number of operations on all types of resources; the number of resource types involved is counted, i.e., the total number of different types of resources interacted with by the user within that time period; the total duration of each type of interaction is counted, accurate to the second; and the number of occurrences of anomalous features is counted, i.e., the number of interactions belonging to the marked anomalous features within that time period. For hourly segments with no interaction, all statistical values ​​are recorded as 0. After parsing, time-series data arranged in chronological order is generated, including time segment identifiers, total number of interactions, number of resource types, duration of each type of interaction, and number of anomalous features, fully reflecting the distribution pattern and trend of user interaction behavior over time.

[0049] Step S254: Based on the user cultural resource interaction time series data, perform user cultural resource interaction specificity time series pattern analysis on the user cultural resource interaction specificity data to generate user cultural resource interaction specificity time series pattern data. In this embodiment of the invention, interaction-specific temporal pattern analysis is performed on interaction-specific data based on user cultural resource interaction time-series data. A Hidden Markov Model (HMM) is employed, with five hidden states representing initial cognitive contact, long-term low-frequency contact, short-term low-frequency contact, short-term high-frequency contact, and long-term high-frequency contact. The model's observations are the number of specific features in each hourly segment. The Baum-Welch algorithm is used to train the model, initializing the state transition probability matrix as uniformly distributed, and the observation probability matrix is ​​set based on the assumption of a normal distribution of the number of specific features. The model is iteratively trained 50 times until the model parameters converge. The convergence criterion is that the change in the likelihood function value is less than 1e-6. After training, the Viterbi algorithm is used to decode the time-series data to obtain the hidden state corresponding to each time-series segment, i.e., the cognitive load level of that time period. The state transition probability of adjacent time-series segments is calculated, which is the number of transitions from state i to state j divided by the total number of occurrences of state i. The generated interaction-specific temporal pattern data includes time segment identifiers, corresponding hidden states, state transition probability matrices, and the average number of unique features for each state, clearly presenting the patterns of user interaction specificity changing over time and the transition rules of cognitive load states.

[0050] Step S255: Based on the user cultural resource interaction-specific time series pattern data, perform user cultural resource cognitive load assessment processing to generate user cultural resource cognitive load assessment data.

[0051] In this embodiment of the invention, cognitive load assessment is performed based on user cultural resource interaction-specific temporal pattern data. First, the correspondence between five hidden states and cognitive load levels is clarified: long-duration high-frequency contact and short-duration high-frequency contact are classified as high-load states; initial cognitive contact is classified as a medium-load state; and long-duration low-frequency contact and short-duration low-frequency contact are classified as low-load states. The total duration of each load state is calculated by multiplying the number of time segments corresponding to each state by the temporal interval. The total duration of high-load states is divided by the total analysis time to obtain the high-load percentage. The state transition frequency is calculated, which is the total number of times the state differs between adjacent time segments divided by the total analysis time. A higher frequency indicates more frequent fluctuations in load states. The cognitive load score is obtained by multiplying the high-load percentage by the state transition frequency and then by 5. The score range is limited to 0 to 5, with a higher score indicating a higher overall cognitive load. High-load periods are identified as those where two or more consecutive time segments are in a high-load state. The start and end times of each period are recorded, and the average number of unique features within that period is calculated by dividing the total number of unique features by the number of time segments. The generated user cultural resource cognitive load assessment data includes cognitive load score, high-load percentage, state transition frequency, a list of high-load periods, and the average number of unique features during high-load periods. This comprehensively quantifies the user's cognitive load level, distribution patterns, fluctuations, and correlation with interaction specificity during the cultural resource learning process.

[0052] Furthermore, step S3 includes the following steps: Step S31: Based on user cultural resource cognitive behavior data and cultural resource cognitive characteristic clustering data of cultural resource cognitive graph, perform user cultural resource interest cluster analysis to generate user cultural resource interest cluster data. In this embodiment of the invention, user cultural resource interest cluster analysis is performed based on user cultural resource cognitive behavior data and cultural resource cognitive characteristic clustering data from a cultural resource cognitive graph, using the K-means clustering algorithm. User cultural resource cognitive behavior data provides behavioral indicators such as interaction frequency and average dwell time. The interaction intensity is obtained by multiplying the interaction frequency by the average dwell time, serving as a weighting basis for resource importance. The cultural resource cognitive characteristic clustering data provides feature vectors for each resource, including dimensions such as topic entities, core attributes, and sentiment tendencies. The resource feature vectors weighted by interaction intensity are used as clustering input. Cosine similarity is used as the distance metric; similarity is determined by calculating the cosine of the angle between two vectors, with values ​​closer to 1 indicating greater similarity in resource cognitive characteristics. The optimal number of clusters is determined by the silhouette coefficient. The silhouette coefficient is calculated by taking the ratio of the difference between a sample and the average distance between it and the nearest dissimilar sample, and the maximum value of the difference. Clusters with a coefficient higher than 0.7 are determined as the final number of clusters. During the clustering process, cultural resources with high-frequency user interaction and similar characteristics are grouped into the same interest cluster. Each cluster includes core resource identifiers (the 3 resources with the highest interaction intensity), the total number of resources in the cluster, the average interaction intensity (the average of the interaction intensity of all resources in the cluster), and core cognitive characteristic tags (based on the high attention characteristics shared by resources in the cluster). The generated user cultural resource interest cluster data integrates the above information to clearly present the core interest directions of users in the field of red culture and the interaction depth of each direction.

[0053] Step S32: Based on the user's cultural resource interest cluster data, perform derivative association data analysis on the cultural resource cognitive map to generate interest cluster derivative association data; In this embodiment of the invention, the derivative association data of user cultural resource interest clusters is analyzed based on user cultural resource cognitive graph. Starting from the core resource node in the interest cluster, the association nodes in the cultural resource cognitive graph are traversed. The traversal scope covers first-degree and second-degree association nodes. First-degree association nodes refer to nodes directly connected to the core node, and second-degree association nodes refer to nodes indirectly connected to the core node through one intermediate node. The association relationship is determined based on the weighted cognitive path in the cultural resource cognitive graph. The path weight must be higher than 0.4; associations with a value lower than this are not included in the analysis due to insufficient strength. For the traversed derivative nodes, their feature information is extracted, including topic entity tags, core attribute sets, association paths with core cluster nodes (including intermediate node identifiers), and total path weight (the product of the weights of each path segment). At the same time, the frequency of each derivative node in the user's historical interactions is counted. Nodes with a frequency of 0 are marked as "non-interactive derivative nodes," and nodes with a frequency greater than 0 are marked as "interactive derivative nodes," and the interaction intensity is recorded. The generated interest cluster-derived association data includes derivative node identifiers, feature information, association path details, total path weight, interaction status and frequency, comprehensively capturing the potential extension associations of core interest clusters to other cultural resources.

[0054] Step S33: Perform interest intensity assessment on the interest clusters and derived data of the interest clusters to generate interest cluster derived intensity assessment data; In this embodiment of the invention, the interest intensity assessment of interest clusters and derived data is performed on the interest clusters and derived data, which is divided into two parts: mutual information analysis and propagation characteristic analysis. Mutual information analysis calculates the mutual information value between derived nodes and core cluster nodes. This value is obtained by dividing the joint probability of the two nodes (the probability of both appearing in the user interaction sequence) by the marginal probability (the probability of each node appearing individually). A higher value indicates a stronger association between the two nodes in user interaction. Propagation characteristic analysis uses the PageRank algorithm, with the core cluster node as the initial seed node, set to an initial weight of 1, and other nodes set to an initial weight of 0. The algorithm iterates 100 times, with a damping coefficient of 0.85, representing the probability of a node jumping to any adjacent node. The final stable weight is the propagation characteristic value of the derived node, reflecting its influence in interest diffusion. The mutual information value and the propagation characteristic value are weighted and summed in a 6:4 ratio to obtain the interest intensity assessment value, which is limited to between 0 and 1. A higher value indicates a stronger potential attraction of the derived node to the user. The generated interest cluster derivative strength assessment data includes derivative node identifiers, mutual information values, propagation characteristic values, and comprehensive interest strength assessment values, which quantitatively characterize the correlation strength between core interest clusters and derivative nodes, as well as the potential attractiveness of derivative nodes.

[0055] Step S34: Analyze the evolutionary characteristics of user interest content increments in user cultural resource interest cluster data using interest cluster derivative intensity assessment data, and generate user cultural resource interest evolutionary characteristic data.

[0056] In this embodiment of the invention, the evolutionary characteristics of user interest content increments are analyzed using interest cluster derivative intensity assessment data. Historical data for four consecutive windows are selected, with a weekly time window. For each interest cluster, the core intensity change rate is calculated, which is the average interaction intensity of the core resources in the current window minus the intensity of the previous window, and then divided by the intensity of the previous window. Simultaneously, the derivative intensity change rate is calculated by counting the number of derivative nodes whose interest intensity assessment value increases in the current window, and dividing by the total number of derivative nodes to obtain the growth ratio. Derivative nodes with positive intensity change rates are captured, and their characteristics are analyzed for correlation with the core cluster characteristics (based on cosine similarity of feature vectors). Nodes with a correlation higher than 0.6 are classified as "co-directional expansion," and those lower than 0.6 are classified as "an-directional expansion." New derivative nodes added in each window (nodes not appearing in the previous window) are recorded, and their intensity change trends in subsequent windows are tracked, categorized into three types: continuous growth, fluctuation, and decline. The generated user cultural resource interest evolution characteristic data includes the core intensity change trend of each interest cluster, the growth ratio of derivative intensity, the distribution of expansion types (proportion of same direction / dissimilar direction), the characteristics and trend trajectory of newly added derivative nodes, fully presenting the incremental evolution law of user interests from core to derivative content, and reflecting the expansion direction and intensity change dynamics of the interest scope.

[0057] Furthermore, step S33 includes the following steps: Step S331: Perform mutual information analysis on the derived data based on the interest cluster-derived association data to generate interest cluster-derived mutual information data; In this embodiment of the invention, mutual information analysis of derived data is performed based on the related data derived from interest clusters. Taking the core resource nodes in the interest cluster as the benchmark, co-occurrence records of the core node and each derived node in the user's historical interaction sequence are extracted. The interaction sequence is arranged in chronological order, and co-occurrence refers to the consecutive appearance of the core node and the derived node in the same sequence, with an interval of no more than 3 other nodes. The joint occurrence frequency of the core node and the derived node is counted and divided by the total interaction sequence length to obtain the joint probability; the marginal probability of the core node is obtained by counting the number of times the core node appears alone and dividing by the total interaction sequence length; the marginal probability of the derived node is obtained by counting the number of times the derived node appears alone and dividing by the total interaction sequence length. The mutual information value is calculated by dividing the joint probability by the product of the marginal probability of the core node and the marginal probability of the derived node. The larger the value, the stronger the association between the two in the user interaction. The above calculation is repeated for each derived node. The generated mutual information data of interest clusters includes the derived node identifier, the core node identifier, the joint probability, the marginal probability of the core node, the marginal probability of the derived node, and the mutual information value, which accurately reflects the interaction association strength between the derived node and the core interest cluster.

[0058] Step S332: Analyze the interest propagation characteristics of the derived data based on the interest cluster-derived related data, and generate interest cluster-derived propagation characteristic data; In this embodiment of the invention, the interest propagation characteristics of derived data are analyzed based on the related data derived from interest clusters, using the PageRank algorithm. Core resource nodes in the interest clusters are set as initial seed nodes, assigned an initial weight of 1; other derived nodes have an initial weight of 0. The algorithm is based on the association paths in the cultural resource cognitive graph, with path weights serving as the attenuation coefficient for propagation between nodes; that is, the propagation weight from node A to node B is the path weight value. In each iteration, each node distributes its own weight to neighboring nodes according to the path weight ratio, while retaining 30% of its own weight (with a damping coefficient of 0.3), and the remaining 70% participates in the distribution. The iteration process continues for 100 times until the change in node weight is less than 1e-5, reaching stability. The final stable weight of each derived node is its propagation characteristic value; a higher value indicates a stronger influence of the node in the propagation path of the core interest cluster. The generated interest cluster derived propagation characteristic data includes derived node identifiers, initial weights, final propagation characteristic values, number of iterations converging, and the contribution ratio of association paths to propagation (the proportion of weight allocated to each path to the total allocated weight), comprehensively characterizing the propagation ability of derived nodes in interest diffusion.

[0059] Step S333: Based on the mutual information data derived from interest clusters and the propagation characteristic data derived from interest clusters, perform interest intensity assessment processing on interest clusters and derived data to generate interest cluster derivation intensity assessment data.

[0060] In this embodiment of the invention, interest intensity assessment processing of interest clusters and derived data is performed based on mutual information data derived from interest clusters and propagation characteristic data derived from interest clusters, and a weighted fusion rule for the two is determined. The mutual information value reflects the closeness of the interaction between the derived node and the core cluster, while the propagation characteristic value reflects the influence of the derived node in interest propagation. The two are calculated with a fixed weight of 6:4. For each derived node, the mutual information value in the mutual information data derived from the interest cluster and the propagation characteristic value in the propagation characteristic data derived from the interest cluster are calculated to obtain the corresponding comprehensive assessment value. The generated interest cluster derived intensity assessment data includes the derived node identifier, mutual information value, propagation characteristic value, comprehensive interest intensity assessment value, and weight allocation explanation. The quantified comprehensive assessment value intuitively presents the potential interest attraction of the derived node to users, providing a core judgment basis for subsequent interest evolution analysis.

[0061] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1 A detailed flowchart of step S4 is shown below. In this embodiment, step S4 includes the following steps: Step S41: Analyze the set of potential cognitive paths of user interests based on the evolutionary characteristics of user cultural resource interests, and generate the set of potential cognitive paths of user interests data; In this embodiment of the invention, a set analysis of potential cognitive paths of user interests is performed based on user cultural resource interest evolution characteristic data. Using four consecutive weeks as time windows, the core nodes, newly added derivative nodes, and distribution of expansion types of interest clusters within each window are extracted. The core intensity change trend (e.g., continuous growth or fluctuation), the proportion of unidirectional / heterogeneous expansion, the feature vectors and trend trajectories of newly added derivative nodes (continuous growth, fluctuation, decline) are obtained from the user cultural resource interest evolution characteristic data to construct a historical cognitive path sequence. The sequence records the node jump order and corresponding interaction intensity changes in chronological order. Based on the associated paths of the cultural resource cognitive graph, potential extension nodes are screened starting from the current core node: unidirectional expansion corresponds to derivative nodes with a cosine similarity higher than 0.6 to the feature vector of the core node, while heterogeneous expansion corresponds to derivative nodes with a similarity lower than 0.6. Each potential path is limited to containing three nodes (starting point, intermediate node, and ending point), and the interest intensity evaluation value of the nodes in the path must decrease by no more than 20% with each successive node (ensuring path continuity). The probability of each potential path appearing in the historical sequence is calculated by dividing the number of matches between the path and similar paths by the total number of historical sequences. Paths with a probability higher than 15% are included in the set. The generated set of potential cognitive paths for user interests includes path identifiers, starting nodes, intermediate nodes, ending nodes, path length, extension type (same direction / opposite direction), historical occurrence probability, and interest intensity evaluation values ​​for each node, fully presenting all potential cognitive path options that users may extend along the direction of interest evolution.

[0062] Step S42: Perform a correlation analysis on the user's interest intensity and cognitive difficulty in the set of potential cognitive paths of user interests, and generate user interest intensity-cognitive difficulty correlation data; In this embodiment of the invention, a correlation analysis is performed on the user's interest intensity and cognitive difficulty in the potential cognitive path set data. The interest intensity is the comprehensive interest intensity assessment value of each node in the path, and the cognitive difficulty is calculated by the node feature complexity, including the number of entities associated with the node and the hierarchy of attribute relationships (such as the number of "whole-part" nesting levels). For each potential path, the Pearson correlation coefficient between the mean interest intensity and the mean cognitive difficulty on the path is calculated. A positive coefficient indicates that the interest intensity increases with the increase of difficulty, and a negative coefficient indicates the opposite. Five intensity-difficulty intervals are divided: interest intensity [0-0.2] corresponds to difficulty [0-0.2] which is low-low interval, [0.2-0.4] corresponds to [0.2-0.4] which is medium-low to medium-low interval, and so on down to [0.8-1] which corresponds to [0.8-1] which is high-high interval. The proportion of nodes in each interval for each path is counted. The generated user interest intensity-cognitive difficulty correlation data includes path identifiers, correlation coefficients, node proportions in each interval, mean interest intensity, and mean cognitive difficulty, clearly presenting the correlation pattern between interest intensity and cognitive difficulty in potential paths.

[0063] Step S43: Design the user cultural resource interest exploration strategy space based on the correlation data of user interest intensity and cognitive difficulty, and generate user cultural resource interest exploration strategy space data; In this embodiment of the invention, a strategy space for exploring user cultural resources interests is designed based on the correlation data between user interest intensity and cognitive difficulty. The strategy space is divided into dimensions according to the type of intensity-difficulty correlation: paths with positive correlation coefficients correspond to "challenging strategies," those with negative correlation coefficients correspond to "comfortable strategies," and those close to 0 correspond to "balanced strategies." Each strategy includes path selection rules, resource type ratios, and difficulty progression: the challenging strategy prioritizes paths where the proportion of nodes in the high-high interval exceeds 40%, with documentaries accounting for no less than 50% of the resource types, and the difficulty increase between adjacent nodes not exceeding 15%; the comfortable strategy prioritizes paths where the proportion of nodes in the high-low interval exceeds 40%, with documentary images accounting for no less than 50%, and difficulty fluctuations controlled within 5%; the balanced strategy selects paths with even distribution across intervals, with the three resource types in a 3:3:4 ratio, and the difficulty progression set to 10%. The strategy space also includes strategy triggering conditions: the comfortable strategy is triggered when the user experiences three consecutive high-load interactions, the challenging strategy is triggered when the user experiences three consecutive low-load interactions, and the balanced strategy is triggered otherwise. The generated user cultural resource interest exploration strategy space data includes strategy type, path selection rules, resource type ratio, difficulty progression, and triggering conditions, forming a structured exploration strategy system.

[0064] Step S44: Analyze the gradient distribution of the utility of interest exploration strategies based on the spatial data of user cultural resource interest exploration strategies, and generate gradient distribution data of the utility of interest exploration strategies. In this embodiment of the invention, the utility gradient distribution analysis of interest exploration strategies is performed based on the spatial data of user cultural resource interest exploration strategies. Utility is comprehensively measured by three interactive feedback indicators: path completion rate is the ratio of the number of actually completed path nodes to the total number of path nodes; average dwell time growth rate is the result of subtracting the historical average dwell time from the current average dwell time and dividing by the historical average dwell time; and note-taking operation occurrence rate is the number of nodes where note-taking operations occurred divided by the total number of path nodes. The three indicators are weighted and summed with a fixed weight of 4:3:3 to obtain the strategy utility value, and the result is normalized to 0 to 1, with higher values ​​indicating better utility. Strategy execution data from the past three time windows are selected and grouped by strategy type (challenging, comfortable, and balanced). Within each group, utility values ​​are calculated based on the interest intensity-cognitive difficulty range (interest intensity 0-0.2 to 0.8-1, cognitive difficulty 0-1 to 4-5), forming a 5×5 utility matrix. Contour lines of utility values ​​were plotted with cognitive difficulty on the horizontal axis and interest intensity on the vertical axis. A difference in utility between adjacent intervals greater than 0.2 was defined as a steep gradient, and less than 0.1 as a gentle gradient. The location and gradient percentage of high-utility intervals (utility value ≥ 0.7) were statistically analyzed for each strategy type. For example, the steep gradient percentage for the challenging strategy reached 60% in the interest intensity 0.8-1 and cognitive difficulty 4-5 intervals, while the gentle gradient percentage for the comfortable strategy reached 70% in the interest intensity 0.8-1 and cognitive difficulty 0-1 intervals. The generated interest exploration strategy utility gradient distribution data includes strategy type, utility matrix, contour line plotting parameters, steep / gentle gradient percentage, and high-utility interval coordinates, quantitatively presenting the utility variation patterns of different strategies.

[0065] Step S45: Based on the spatial data of user cultural resource interest exploration strategies and the gradient distribution data of interest exploration strategy utility, establish the learning mapping relationship between interest exploration and utility, and generate a user cultural resource interest exploration learning model; In this embodiment of the invention, a learning mapping relationship between interest exploration and utility is established based on user cultural resource interest exploration strategy space data and interest exploration strategy utility gradient distribution data, using a multiple linear regression model. The model input consists of feature parameters from the strategy space data: resource type ratio, difficulty progression, state parameters corresponding to triggering conditions (such as load state values), and the high-utility interval position from the gradient distribution data; the output is the strategy utility value. The model is trained using historical data from the past five time windows. During training, the mean squared error between the predicted and actual utility values ​​is calculated, and the regression coefficients are iteratively adjusted until the error is less than 0.05. The model includes an input layer (6 feature parameters), an output layer (1 utility value), and a linear mapping formula: Utility value = (each feature parameter × corresponding coefficient) + constant term. The mapping relationship also includes the correspondence rules between strategy type and optimal parameter combination, such as the highest weight for "literature text proportion" in the coefficient matrix for challenging strategies. The generated user cultural resource interest exploration learning model includes an input feature set, output index, regression coefficient matrix, error range, and strategy-parameter correspondence rules, achieving a quantitative mapping from exploration strategy to utility result.

[0066] Step S46: Use the user cultural resource interest exploration learning model to perform user cultural resource learning interest behavior prediction processing on the user interest potential cognitive path set data, and generate user cultural resource learning interest behavior prediction data.

[0067] In this embodiment of the invention, a user cultural resource interest exploration learning model is used to predict user cultural resource learning interest behavior based on the set of potential cognitive paths for user interests. The node features (interest intensity, cognitive difficulty) of each potential path are extracted, along with matching strategy parameters (resource type ratio, difficulty progression) in the strategy space. These are then input into the model to obtain the predicted utility value for each path. The paths are sorted from highest to lowest predicted utility value, and the top three paths are selected as core predicted paths. For each core path, user interaction behavior indicators are predicted: the path completion probability is predicted utility value × 0.8 plus historical completion rate × 0.2; the predicted dwell time for each node is the historical average dwell time multiplied by (1 plus predicted utility value × 0.3); among the predicted operation types, when the predicted utility value is higher than 0.7, the probability of note-taking is increased by 50% compared to the historical level. Simultaneously, the selection probability of each path is calculated as the proportion of the predicted utility value of that path to the total predicted utility value of the three core paths. The generated user cultural resource learning interest behavior prediction data includes prediction path identifiers, path completion probability, predicted dwell time at each node, prediction operation type and probability, and path selection probability, accurately outputting the user's future interest behavior tendencies in cultural resource learning.

[0068] Furthermore, step S42 includes the following steps: Step S421: Perform cultural resource cognitive information entropy analysis on the cultural resource cognitive map to generate cultural resource cognitive information entropy data; In this embodiment of the invention, a cognitive information entropy analysis is performed on the cultural resource cognitive map. Information entropy is calculated from two dimensions: node attributes and relationships, to quantify the complexity of resource cognition. Node attributes include three fixed dimensions: topic entity, core attribute, and description depth. The attribute entropy of each node is calculated by first statistically analyzing the distribution probability of each category under each dimension (the number of times a category appears in the node attributes divided by the total number of categories in that dimension), then calculating the entropy value of each dimension, and finally weighting and summing them according to the weights of topic entity, core attribute, and description depth. A higher value indicates a more dispersed and complex node attribute information. The relationship dimension includes three types: causal relationship, subordinate relationship, and parallel relationship. Relationship entropy is calculated by statistically analyzing the proportion of each relationship type in all relationships of a node, and also using the information entropy formula. A higher value indicates a more diverse range of relationship types. The combined cognitive information entropy is the accumulation of attribute entropy and relationship entropy to generate cultural resource cognitive information entropy data.

[0069] Step S422: Analyze the difficulty of users' cultural resource cognition based on user cultural resource cognitive behavior data and cultural resource cognitive information entropy data, and generate user cultural resource cognition difficulty data; In this embodiment of the invention, the cognitive difficulty of user cultural resources is analyzed based on user cultural resource cognitive behavior data and cultural resource cognitive information entropy data. The cognitive difficulty is quantified by combining the complexity of the resource itself with user interaction performance. User cultural resource cognitive behavior data provides three core indicators: average dwell time (total dwell time at a node divided by the number of interactions), reflecting the time investment users make in understanding the resource; note-taking frequency (number of times notes are added to a node divided by the number of interactions), reflecting the user's need to actively organize information; and repeated visits (total number of visits to a node divided by the total number of interactions), reflecting the difficulty users face in understanding the resource. These three indicators are standardized: each indicator value is divided by the mean of the corresponding indicator for all users in the system to obtain a standardized value. Then, a weighted sum is calculated based on the weights of average dwell time, note-taking frequency, and repeated visits to obtain the behavioral complexity value. A higher value indicates greater user interaction investment at that node, indirectly reflecting the cognitive challenge. The cognitive difficulty value is calculated by accumulating the cognitive information entropy and behavioral complexity values, generating user cultural resource cognitive difficulty data.

[0070] Step S423: Based on the interest cluster derivative intensity assessment data and the user cultural resource cognitive difficulty data of the user interest potential cognitive path set data, perform a correlation analysis on the user interest intensity and cognitive difficulty of the user cultural resource cognitive path set data to generate user interest intensity-cognitive difficulty correlation data.

[0071] In this embodiment of the invention, based on the interest cluster derivative intensity assessment data and the user cultural resource cognitive difficulty data of the user interest potential cognitive path set data, a correlation analysis is performed on the user interest potential cognitive path set data to reveal the intrinsic relationship between interest intensity and cognitive difficulty. Each potential path contains three nodes: a starting point, a middle point, and an ending point. The comprehensive interest intensity assessment value (between 0 and 1) for each node is extracted from the interest cluster derivative intensity assessment data, and the cognitive difficulty value (between 0 and 5) for each node is extracted from the user cultural resource cognitive difficulty data. For each path, the mean interest intensity is calculated by summing the comprehensive interest intensity assessment values ​​of the three nodes and dividing by 3; the mean cognitive difficulty is calculated by summing the cognitive difficulty values ​​of the three nodes and dividing by 3. The correlation between the two was calculated using the Pearson correlation coefficient formula: the numerator is the covariance (the difference between the interest intensity of each node and the mean interest intensity of the path, multiplied by the difference between the cognitive difficulty of the node and the mean cognitive difficulty of the path, summed and divided by the number of nodes); the denominator is the product of the standard deviation of interest intensity and the standard deviation of cognitive difficulty (the standard deviation is the square root of the sum of the squares of the differences between each node's value and the mean, divided by the number of nodes). The coefficient ranges from -1 to 1, with positive numbers indicating that interest intensity increases with increasing difficulty, and negative numbers indicating that it decreases with increasing difficulty. Interest intensity was divided into five continuous intervals: 0-0.2, 0.2-0.4, 0.4-0.6, 0.6-0.8, and 0.8-1; cognitive difficulty was divided into five continuous intervals: 0-1, 1-2, 2-3, 3-4, and 4-5. The proportion of nodes in each path falling into each intensity-difficulty interval was counted out of the total number of nodes in that path, generating user interest intensity-cognitive difficulty correlation data.

[0072] This specification also provides a user learning interest and behavior analysis system for performing the user learning interest and behavior analysis method described above. The user learning interest and behavior analysis system includes: The cultural resource analysis module is used to acquire cultural resource data from the cultural resource system; and to construct a cognitive map of cultural resource associations based on the cultural resource data, thereby generating a cultural resource cognitive map. The User Cultural Resource Cognitive Behavior Analysis Module is used to obtain user cultural resource interaction logs from the cultural resource system; and to perform user cultural resource cognitive behavior analysis based on the user cultural resource interaction logs and cultural resource cognitive maps to generate user cultural resource cognitive behavior data. The User Cultural Resource Interest Evolution Module is used to analyze the evolutionary characteristics of user interest content increments based on user cultural resource cognitive behavior data, and generate user cultural resource interest evolution characteristic data. The user learning interest and behavior prediction module is used to analyze the set of potential cognitive paths of user interests based on the evolutionary characteristics of user cultural resource interests, and generate the set of potential cognitive paths of user interests data; and to perform learning interest and behavior prediction processing on the set of potential cognitive paths of user interests data to generate learning interest and behavior prediction data of user cultural resources.

[0073] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0074] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A user learning interest behavior analysis method based on, characterized by, The method comprises the following steps: Step S1: obtaining cultural resource data of a cultural resource system; constructing a cognitive map of cultural resource association according to the cultural resource data to generate a cultural resource cognitive map; Step S2: obtaining a user cultural resource interaction log of the cultural resource system; analyzing user cultural resource cognitive behaviors according to the user cultural resource interaction log and the cultural resource cognitive map to generate user cultural resource cognitive behavior data; Step S3: analyzing evolution characteristics of user interest content increments according to the user cultural resource cognitive behavior data to generate user cultural resource interest evolution characteristic data; Step S4: analyzing a user interest potential cognitive path set according to the user cultural resource interest evolution characteristic data to generate user interest potential cognitive path set data; and performing learning interest behavior prediction processing on the user interest potential cognitive path set data to generate user cultural resource learning interest behavior prediction data.

2. The user learning interest behavior analysis method of claim 1, wherein, Step S1 comprises the following steps: Step S11: obtaining cultural resource data of a cultural resource system; Step S12: performing cultural resource text analysis processing on the cultural resource data to generate cultural resource text analysis data; Step S13: performing multi-dimensional semantic feature analysis processing on the cultural resource text analysis data to generate cultural resource multi-dimensional semantic feature data; Step S14: performing cultural resource attention cognitive characteristic analysis on the cultural resource multi-dimensional semantic feature data based on a preset attention mechanism to generate cultural resource attention cognitive characteristic data; Step S15: performing cultural resource cognitive characteristic clustering analysis on the cultural resource attention cognitive characteristic data to generate cultural resource cognitive characteristic clustering data; Step S16: designing cultural resource cognitive nodes according to the cultural resource cognitive characteristic clustering data, and performing cultural resource cognitive path analysis on the cultural resource data and the cultural resource cognitive characteristic clustering data to generate cultural resource cognitive path data; Step S17: constructing a cognitive map of cultural resource association according to the cultural resource cognitive nodes and the cultural resource cognitive path data to generate a cultural resource cognitive map.

3. The user learning interest behavior analysis method of claim 2, wherein, Step S13 comprises the following steps: Step S131: performing cultural resource theme entity feature analysis on the cultural resource text analysis data to generate cultural resource theme entity feature data; Step S132: performing cultural resource entity attribute analysis on the cultural resource text analysis data to generate cultural resource entity attribute data; Step S133: performing entity attribute semantic dependency relationship analysis on the cultural resource entity attribute data to generate entity attribute semantic dependency relationship data, and performing cultural resource attribute relationship feature analysis on the entity attribute semantic dependency relationship data to generate cultural resource attribute relationship feature data; Step S134: performing cultural resource sentiment associated entity extraction on the cultural resource entity attribute data to obtain cultural resource sentiment associated entity data; Step S135: performing cultural resource sentiment semantic feature analysis on the cultural resource text analysis data based on the cultural resource sentiment associated entity data to generate cultural resource sentiment semantic feature data; Step S136: Perform multi-dimensional semantic feature aggregation processing on the cultural resource theme entity feature data, the cultural resource attribute relationship feature data, and the cultural resource sentiment semantic feature data to generate cultural resource multi-dimensional semantic feature data.

4. The user learning interest behavior analysis method of claim 1, wherein, Step S2 includes the following steps: Step S21: Obtain a user cultural resource interaction log of a cultural resource system; Step S22: Map the user cultural resource interaction log to a cultural resource cognitive graph to perform preliminary user cultural resource cognitive state analysis and generate preliminary user cultural resource cognitive state data; Step S23: Perform user cultural resource interaction portrait analysis based on the user cultural resource interaction log to generate user cultural resource interaction portrait data; Step S24: Perform user resource cognitive state change characteristic analysis on the preliminary user cultural resource cognitive state data based on the user cultural resource interaction portrait data to generate user resource cognitive state change characteristic data; Step S25: Perform user cultural resource cognitive load evaluation processing based on the user cultural resource interaction portrait data to generate user cultural resource cognitive load evaluation data; Step S26: Perform user cultural resource cognitive behavior analysis based on the user resource cognitive state change characteristic data and the user cultural resource cognitive load evaluation data to generate user cultural resource cognitive behavior data.

5. The user learning interest behavior analysis method of claim 1, wherein, Step S25 includes the following steps: Step S251: Perform user cultural resource type interaction feature analysis based on the user cultural resource interaction portrait data to generate user cultural resource type interaction feature data; Step S252: Perform user cultural resource interaction specificity analysis based on the user cultural resource type interaction feature data to generate user cultural resource interaction specificity data; Step S253: Perform interaction time sequence data analysis based on the user cultural resource interaction portrait data to generate user cultural resource interaction time sequence data; Step S254: Perform user cultural resource interaction specificity time sequence mode analysis on the user cultural resource interaction specificity data based on the user cultural resource interaction time sequence data to generate user cultural resource interaction specificity time sequence mode data; Step S255: Perform user cultural resource cognitive load evaluation processing based on the user cultural resource interaction specificity time sequence mode data to generate user cultural resource cognitive load evaluation data.

6. The user learning interest behavior analysis method based on claim 2, characterized in that, Step S3 includes the following steps: Step S31: Perform user cultural resource interest cluster analysis based on the user cultural resource cognitive behavior data and cultural resource cognitive characteristic clustering data of the cultural resource cognitive graph to generate user cultural resource interest cluster data; Step S32: Perform user cultural resource interest cluster derivative association data analysis on the cultural resource cognitive graph based on the user cultural resource interest cluster data to generate interest cluster derivative association data; Step S33: Perform interest intensity evaluation processing on the interest cluster and derivative data based on the interest cluster derivative association data to generate interest cluster derivative intensity evaluation data; Step S34: Perform user interest content increment evolution feature analysis on the user cultural resource interest cluster data based on the interest cluster derivative intensity evaluation data to generate user cultural resource interest evolution feature data.

7. The user learning interest behavior analysis method based on claim 6, characterized in that, Step S33 includes the following steps: Step S331: Derivative data mutual information analysis is performed according to the interest cluster derived correlation data, and interest cluster derived mutual information data is generated; Step S332: Interest propagation characteristic analysis of derivative data is performed according to the interest cluster derived correlation data, and interest cluster derived propagation characteristic data is generated; Step S333: Interest intensity evaluation processing of the interest cluster and the derivative data is performed based on the interest cluster derived mutual information data and the interest cluster derived propagation characteristic data, and interest cluster derived intensity evaluation data is generated.

8. The user learning interest behavior analysis method based on claim 6, characterized in that, Step S4 includes the following steps: Step S41: User interest potential cognitive path set analysis is performed according to the user cultural resource interest evolution characteristic data, and user interest potential cognitive path set data is generated; Step S42: Interest intensity and cognitive difficulty correlation analysis of the user cultural resource is performed on the user interest potential cognitive path set data, and user interest intensity-cognitive difficulty correlation data is generated; Step S43: User cultural resource interest exploration strategy space design is performed according to the user interest intensity-cognitive difficulty correlation data, and user cultural resource interest exploration strategy space data is generated; Step S44: Interest exploration strategy utility gradient distribution analysis is performed according to the user cultural resource interest exploration strategy space data, and interest exploration strategy utility gradient distribution data is generated; Step S45: Interest exploration and utility learning mapping relationship establishment is performed based on the user cultural resource interest exploration strategy space data and the interest exploration strategy utility gradient distribution data, and a user cultural resource interest exploration learning model is generated; Step S46: User cultural resource learning interest behavior prediction processing is performed on the user interest potential cognitive path set data by using the user cultural resource interest exploration learning model, and user cultural resource learning interest behavior prediction data is generated.

9. The user learning interest behavior analysis method based on claim 8, characterized in that, Step S42 includes the following steps: Step S421: Cultural resource cognitive information entropy analysis is performed on the cultural resource cognitive graph, and cultural resource cognitive information entropy data is generated; Step S422: User cultural resource cognitive difficulty analysis is performed according to the user cultural resource cognitive behavior data and the cultural resource cognitive information entropy data, and user cultural resource cognitive difficulty data is generated; Step S423: Interest intensity and cognitive difficulty correlation analysis of the user cultural resource is performed on the user interest potential cognitive path set data according to the interest cluster derived intensity evaluation data of the user interest potential cognitive path set data and the user cultural resource cognitive difficulty data, and user interest intensity-cognitive difficulty correlation data is generated.

10. A user learning interest behavior analysis system based on, characterized by, The user learning interest behavior analysis system is used for performing the user learning interest behavior analysis method, and includes: A cultural resource analysis module, configured to acquire cultural resource data of a cultural resource system; perform cultural resource correlation cognitive graph construction according to the cultural resource data, and generate cultural resource cognitive graph data; A user cultural resource cognitive behavior analysis module, configured to acquire user cultural resource interaction log of the cultural resource system; perform user cultural resource cognitive behavior analysis according to the user cultural resource interaction log and the cultural resource cognitive graph, and generate user cultural resource cognitive behavior data; A user cultural resource cognitive behavior analysis module, configured to acquire user cultural resource interaction log of the cultural resource system; perform user cultural resource cognitive behavior analysis according to the user cultural resource interaction log and the cultural resource cognitive graph, and generate user cultural resource cognitive behavior data; The user cultural resource interest evolution module is configured to perform evolution feature analysis on the user interest content increment according to the user cultural resource cognitive behavior data, and generate user cultural resource interest evolution feature data. The user learning interest behavior prediction module is configured to perform analysis on a user interest potential cognitive path set according to the user cultural resource interest evolution feature data, and generate user interest potential cognitive path set data; and perform learning interest behavior prediction processing on the user cultural resource according to the user interest potential cognitive path set data, and generate user cultural resource learning interest behavior prediction data.

Images