Learning performance evaluation driven teaching management method constructed based on capability atlas

Abstract

The invention provides a learning performance evaluation-driven teaching management method constructed based on a capability graph. The method comprises the following steps: S1, constructing a multi-dimensional capability graph; s2, a step of operating a multi-modal data acquisition system; s3, performing a dynamic capability value calculation model; s4, a step of constructing a real-time capability early warning system; s5, generating a self-adaptive teaching strategy; s6, a step of carrying out interdisciplinary ability association analysis; s7, generating a personalized learning path; s8, dynamically evaluating the teaching effect; and S9, dynamically optimizing the teaching resources. According to the method, the core problems of data lag, inaccuracy in intervention and one-sided evaluation in traditional education management are solved by constructing a dynamic capability evaluation model, and a quantifiable decision support system is provided for precise teaching. The innovation of the method is that differential equation modeling and reinforcement learning are combined, and continuous optimization and adaptive adjustment of the education process are realized.

Description

Technical Field

[0001] The present invention specifically relates to a teaching management method driven by learning performance evaluation based on capability map construction. Background Art

[0002] The current education evaluation system primarily relies on standardized test scores, which account for approximately 80% of the weight of teaching management decisions, combined with subjective teacher evaluations and student peer evaluations. This traditional model suffers from three core issues: severe data silos, preventing most schools from integrating cross-disciplinary competency data; significant evaluation lags, with an average data processing delay of several weeks; and a lack of dynamic adjustment capabilities, with only a small number of teachers able to adjust teaching strategies based on real-time data. In summary, this application proposes a learning performance evaluation-driven teaching management method based on competency maps to address these issues. Summary of the Invention

[0003] The purpose of the present invention is to address the deficiencies of the existing technology and provide a teaching management method driven by learning performance evaluation based on a capability map. The teaching management method driven by learning performance evaluation based on a capability map can well solve the above problems.

[0004] To achieve the above requirements, the present invention adopts a technical solution: providing a teaching management method driven by learning performance evaluation based on capability map construction, the teaching management method driven by learning performance evaluation based on capability map construction includes the following steps:

[0005] S1: Steps for constructing a multidimensional capability map, including determining capability dimensions and indicator systems, establishing capability association models, developing map visualization tools, and developing capability assessment tools;

[0006] S2: Steps for implementing a multimodal data acquisition system, including deploying intelligent sensing devices, implementing data cleaning and fusion, and building a data storage architecture;

[0007] S3: Steps for developing a dynamic capability value calculation model, including establishing a differential equation model, implementing capability prediction, and generating a capability development report;

[0008] S4: Steps for building a real-time capability warning system, including building a real-time data stream processing pipeline, designing a multi-level warning rule engine, and implementing a hierarchical response mechanism;

[0009] S5: Steps for generating adaptive teaching strategies, including building a teaching action feature library, implementing reinforcement learning strategy optimization, and generating explainable teaching plans;

[0010] S6: Steps for conducting cross-disciplinary competence correlation analysis, including constructing a knowledge transfer map, mining implicit competence correlations, and designing collaborative training programs;

[0011] S7: Steps for generating personalized learning paths, including building multi-dimensional user profiles, implementing path planning algorithms, and generating dynamic adjustment plans;

[0012] S8: Steps for conducting dynamic evaluation of teaching effectiveness, including constructing multimodal evaluation indicators, implementing a comprehensive evaluation model, and conducting teaching effectiveness diagnosis;

[0013] S9: Steps for dynamic optimization of teaching resources, including building a resource evaluation system, implementing intelligent recommendation strategies, and building a resource evolution mechanism.

[0014] The advantages of this teaching management method driven by learning performance evaluation based on the capability map are as follows:

[0015] Dynamically track capability development: The differential equation model enables continuous calculation of capability values, significantly increasing the frequency of data updates compared to traditional periodic testing.

[0016] Precision teaching intervention: The reinforcement learning model reduces the time required to adjust teaching strategies from weeks to minutes.

[0017] Interdisciplinary ability training: GNN analysis discovers implicit ability association paths and optimizes curriculum design.

[0018] Personalized learning: The path planning algorithm greatly improves learning efficiency and time utilization.

[0019] Process evaluation: The TOPSIS index reflects the development trend of capabilities more accurately than the traditional percentage system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The same reference numerals are used in these drawings to represent the same or similar parts. The exemplary embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0021] Figure 1 The flowchart of the teaching management method driven by learning performance evaluation based on the capability map construction according to one embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0022] In order to make the objectives, technical solutions and advantages of this application clearer, this application is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] In the following description, references to "one embodiment," "an embodiment," "an example," "an example," etc. indicate that the embodiment or example described may include certain features, structures, characteristics, properties, elements, or limitations, but not every embodiment or example necessarily includes the certain features, structures, characteristics, properties, elements, or limitations. In addition, repeated use of the phrase "according to one embodiment of the present application" may refer to the same embodiment, but does not necessarily refer to the same embodiment.

[0024] For the sake of simplicity, certain technical features well known to those skilled in the art are omitted in the following description.

[0025] According to one embodiment of the present application, a teaching management method driven by learning performance evaluation based on capability map is provided, such as Figure 1 As shown, a teaching management method driven by learning performance evaluation based on capability map construction is characterized by comprising the following steps:

[0026] S1: Steps for constructing a multidimensional capability map, including determining capability dimensions and indicator systems, establishing capability association models, developing map visualization tools, and developing capability assessment tools;

[0027] S2: Steps for implementing a multimodal data acquisition system, including deploying intelligent sensing devices, implementing data cleaning and fusion, and building a data storage architecture;

[0028] S3: Steps for developing a dynamic capability value calculation model, including establishing a differential equation model, implementing capability prediction, and generating a capability development report;

[0029] S4: Steps for building a real-time capability warning system, including building a real-time data stream processing pipeline, designing a multi-level warning rule engine, and implementing a hierarchical response mechanism;

[0030] S5: Steps for generating adaptive teaching strategies, including building a teaching action feature library, implementing reinforcement learning strategy optimization, and generating explainable teaching plans;

[0031] S6: Steps for conducting cross-disciplinary competence correlation analysis, including constructing a knowledge transfer map, mining implicit competence correlations, and designing collaborative training programs;

[0032] S7: Steps for generating personalized learning paths, including building multi-dimensional user profiles, implementing path planning algorithms, and generating dynamic adjustment plans;

[0033] S8: Steps for conducting dynamic evaluation of teaching effectiveness, including constructing multimodal evaluation indicators, implementing a comprehensive evaluation model, and conducting teaching effectiveness diagnosis;

[0034] S9: Steps for dynamic optimization of teaching resources, including building a resource evaluation system, implementing intelligent recommendation strategies, and building a resource evolution mechanism.

[0035] According to one embodiment of the present application, step S1 of the teaching management method driven by learning performance evaluation based on capability map construction specifically includes:

[0036] S11: Determine the competency dimensions and indicator system. Organize an interdisciplinary expert group (including educational psychologists, subject leaders, and data scientists) and use three rounds of Delphi to determine the core competency framework: the first round collects open-ended opinions and forms 128 potential competency items; the second round uses a 5-point Likert scale to assess importance and screens out 42 key competencies; the third round verifies the indicator discrimination and calculates the Cronbach's alpha coefficient for each indicator (required to be >0.7);

[0037] Construct a three-level indicator tree: First-level dimensions: knowledge reserve (K), skill level (S), cognitive literacy (C), and emotional attitude (A); Second-level indicators: For example, the knowledge dimension includes "concept mastery" (K1) and "knowledge transfer ability" (K2); Third-level indicators: detailed into "application of the fundamental theorem of calculus" (K1-3) and "interdisciplinary analogical reasoning" (C2-1);

[0038] Weight calculation: Use the analytic hierarchy process (AHP) to construct a judgment matrix and calculate the weight of each level. For example, the knowledge dimension weight = 0.35, where K1 weight is 0.4, K2 weight is 0.3, and K3 weight is 0.3. A consistency test (CR < 0.1) is performed to ensure the rationality of the weight distribution.

[0039] S12: Establish an ability association model and construct an ability migration map: nodes represent ability items, and edge weights are calculated using historical data (such as the Spearman correlation coefficient). For example, the correlation strength of "spatial imagination ability → physical modeling ability" is 0.72;

[0040] Develop graph visualization tools: Use the Neo4j graph database to store relational data; implement interactive query functions (e.g., input current capability values ​​and recommend related capabilities for priority development);

[0041] Dynamic update mechanism: collect 2000+ pieces of capability development data each semester; use PageRank algorithm to iteratively update node importance scores;

[0042] S13: Develop competency assessment tools and design multimodal assessment scales: Knowledge dimension: Adaptive testing (CAT) is used to dynamically generate 30-50 questions per session; Skill dimension: Through project analysis (e.g., code complexity assessment for programming assignments); Emotional dimension: Using semantic analysis technology to process learning logs (e.g., analyzing the sentiment of "this concept is difficult to understand");

[0043] Construct a calibration dataset: collect sample data from 1,000+ students and annotate their ability development trajectories; calculate the reliability and validity of the assessment tool (Cronbach's α > 0.85, criterion-related validity r > 0.7); deploy an automated scoring engine: use the BERT model for semantic understanding of subjective questions (accuracy 92%); and generate instant feedback reports (average response time < 3 seconds).

[0044] According to one embodiment of the present application, step S2 of the teaching management method driven by learning performance evaluation based on capability map construction specifically includes:

[0045] S21: Deploy intelligent sensing devices and build an IoT data collection network: Deploy eight 4K cameras (sampling rate 30fps) in the classroom to capture facial expressions (such as frown frequency), install millimeter-wave radar to monitor changes in student sitting posture (sampling rate 100Hz), and equip smart bracelets to record physiological indicators (HRV, EDA, sampling rate 25Hz);

[0046] Configure the learning behavior tracking module: the LMS system records resource access paths (such as video playback times and note annotation locations), and the eye tracker tracks gaze heat maps (with an accuracy of 0.5° and a sampling rate of 60Hz);

[0047] Data synchronization mechanism: Use timestamps to align multi-source data (error <1ms), build a data lineage map, and record the collection device and timestamp of each data point;

[0048] S22: Data cleaning and fusion, noise elimination: wavelet threshold denoising (db4 wavelet, threshold selection VisuShrink method) is applied to the camera data, and Kalman filtering is performed on the physiological signal (state transfer matrix Q = 0.01I)

[0049] Feature engineering: extract temporal features (such as the RMSSD value of HRV) and calculate spatial features (such as the entropy value of eye movement heat map);

[0050] Data fusion: Use the federated learning framework (FATE platform) to aggregate data across schools and apply DS evidence theory to fuse multi-source data (initiate manual arbitration when the conflict degree is greater than 0.5);

[0051] S23: Build a data storage architecture and design a tiered storage solution: Raw data layer: HDFS stores raw video / sensor data (retention for 180 days); Feature data layer: MongoDB stores time series features (compression rate 65%); Knowledge data layer: Neo4j stores capability association graphs;

[0052] Implement data lifecycle management: hot data (nearly 30 days) is stored in the SSD cluster, warm data (31-90 days) is migrated to Blu-ray storage, and cold data (>90 days) is transferred to the tape library.

[0053] Security protection mechanism: Apply homomorphic encryption technology to protect sensitive data (CKKS solution) and implement blockchain-based data audit logs.

[0054] According to one embodiment of the present application, step S3 of the teaching management method driven by learning performance evaluation based on capability map construction specifically includes:

[0055] S31: Establish a differential equation model and construct the ability development dynamics equation:

[0056]

[0057] in:

[0058] A(t): ability value at time t (0≤A≤1);

[0059] K: upper limit of ability (set by subject experts, such as K = 0.95 for calculus ability);

[0060] r: individual learning efficiency coefficient (calculated using historical data, range 0.05-0.15);

[0061] t0: Ability inflection point time (when the moment of maximum);

[0062] λ: knowledge internalization coefficient (0.1-0.3);

[0063] I(τ): teaching input intensity function (course difficulty × effective teaching time);

[0064] Parameter calibration, fitting historical data by least squares method (R 2 >0.89); boundary processing, when A(t)>K, it is forced to be normalized to the K value;

[0065] The model integrates biological growth curves and cognitive load theory. When the teaching input intensity I(τ) exceeds a critical value (e.g., I>0.7), the system automatically switches to "deep learning mode" and increases the supply of case analysis resources. Capacity values ​​are calculated in batches at 2:00 a.m. daily, with an error rate within ±3%.

[0066] Develop a numerical solver: Use the fourth-order Runge-Kutta method (step size Δt = 0.5 days) and set a convergence condition (residual < 1e-6). Build an anomaly detection mechanism: When the calculated value deviates from the measured value by >15%, a calibration process is triggered, automatically adjusting the λ parameter (by ±0.05).

[0067] S32: Implement capability prediction and train an LSTM prediction model: Input layer: capability value sequence of the past 30 days, hidden layer: 256 LSTM units (dropout rate 0.2), output layer: capability prediction value of the next 7 days;

[0068] Constructing probabilistic prediction intervals: Using quantile regression to generate 95% confidence intervals, triggering an alert when the lower bound of the prediction falls below a threshold;

[0069] Implement dynamic calibration: fine-tune model parameters using new data every week (learning rate 0.001) and calculate the forecast error MAPE (required to be <8%);

[0070] S33: Generate capability development report and visualize it: draw capability development curve (with 95% confidence interval), generate capability short board radar chart (highlight capability items with a score below 0.6);

[0071] Forming intervention recommendations: matching knowledge gaps (e.g., "integral calculation error → recommending segmented integral training") and calculating resource matching (preferring resources with a cosine similarity > 0.7);

[0072] Automated report generation: Uses a template engine (Jinja2) to dynamically fill in content and output reports in PDF / HTML dual formats (average generation time < 2 minutes).

[0073] According to one embodiment of the present application, step S4 of the teaching management method driven by learning performance evaluation based on capability map construction specifically includes:

[0074] S41: Deploy a distributed stream processing framework (Apache Kafka + Flink):

[0075] Data source access: Convert heterogeneous data from cameras, sensors, and LMS systems into JSON format. Sharding strategy: Hash sharding is performed by student ID (a total of 1024 shards) to ensure data continuity for the same student. Window calculation: Set a sliding window (5 minutes) to calculate the volatility of the statistical ability indicator. The calculation formula is: When volatility > 0.3, an abnormal flag is triggered;

[0076] Implement feature engineering: calculate time domain features: rolling average (window size = 30 minutes), exponentially weighted moving average (α = 0.2), extract frequency domain features: analyze the energy proportion of the low frequency band (0.04-0.15Hz) of heart rate variability (HRV) through FFT transformation

[0077] Build a status monitoring dashboard: Display a real-time heat map of the ability value distribution of each campus, mark the location of abnormal students (GIS map coordinates are accurate to the classroom level), and automatically generate a vector map of the ability development speed (arrow direction indicates improvement / decrease trend)

[0078] S42: Establish a dynamic threshold system: Baseline threshold: Calculate the moving average (EMA=0.7) based on historical data; Adaptive threshold: Automatically adjust the threshold when the class average ability value changes by ±5%; Personalized threshold: Set differentiated parameters for special student groups (such as ADHD)

[0079] Rule logic tree construction:

[0080] A[ability value < 0.2] --> B{lasts for 3 days?};

[0081] B-->|is|C[red alert];

[0082] B-->|No|D[Orange Alert];

[0083] A-->E{Is there any escape behavior?};

[0084] E-->|is|F[red alert];

[0085] Develop a rule verification module: Use a decision tree (CART algorithm) to verify rule coverage (target ≥ 95%); implement conflict resolution: When multiple rules are triggered simultaneously, prioritize the highest confidence rule; and perform monthly rule stress testing (simulating extreme scenarios to verify response speed).

[0086] S43: Build an early warning response knowledge base: a graded intervention plan library, including six basic plans (such as increasing the frequency of practice) and 32 extended plans (such as VR situational training); a learning resource index table, establishing an inverted index by ability item (such as "calculus → piecewise integration → 10 cases");

[0087] Deploy an intelligent decision-making engine: Use a Bayesian network to calculate the posterior probability of intervention measures (e.g., choosing option X will increase capacity by 78%). Conduct a cost-benefit analysis: Calculate the ratio of time cost (Δt) to expected benefit (ΔA) (target ΔA / Δt>0.15).

[0088] Design a closed-loop feedback system: record the implementation status of each intervention (start time, completion rate, student feedback), calculate the strategy effectiveness index (SEI = actual improvement / expected improvement), and automatically trigger the program optimization process when SEI < 0.6.

[0089] According to one embodiment of the present application, step S5 of the teaching management method driven by learning performance evaluation based on capability map construction specifically includes:

[0090] S51: Construct a teaching action feature library and collect teaching intervention samples: record more than 100,000 teaching cases in the past three years (including lesson plans, resources, and student performance changes); annotate key features: teaching method (lecture / exploration), resource type (video / simulation), and duration distribution

[0091] Feature vectorization: Use TF-IDF to extract text features (teaching plan descriptions, learning logs), and perform CNN feature extraction on video resources (ResNet-50 model, outputting 2048-dimensional vectors)

[0092] Constructing a similarity matrix: Calculating cosine similarity (threshold = 0.65) and generating a clustering map of teaching actions (Louvain algorithm divides into 7 strategy clusters);

[0093] S52: Implement reinforcement learning strategy optimization and design the state space: current ability map state (compressed into a 64-dimensional vector), learning environment characteristics (time, device availability, course progress)

[0094] Define the action space: basic actions (such as "recommend micro-course" and "adjust difficulty"), combined actions (up to three levels of nesting are allowed, such as "review first → do experiment → test last");

[0095] Construct the reward function: R(s,a)=0.6ΔA+0.3ΔE+0.1ΔT;

[0096] ΔA: ability improvement rate;

[0097] ΔE: Change in learning efficiency (increment in knowledge points per unit time);

[0098] ΔT: reduction in resource consumption time;

[0099] Training the DQN model: using the Double DQN algorithm to eliminate overestimation, experience replay pool capacity = 1e6, prioritizing samples with high TD error, target network update frequency = 100 steps;

[0100] S53: Generate explainable teaching plans and implement counterfactual analysis of strategies: compare the differences between the current strategy and the historical optimal solution, and mark key decision factors (e.g., "add 3D modeling module due to insufficient spatial imagination");

[0101] Develop a natural language generator: Generate policy instructions based on the BERT model (e.g., "It is recommended to add 2 hours of graphical reasoning training because this student's geometric abstraction ability is 23% lower than that of his peers") and output guidance recommendations that are consistent with educational principles (citing ISTE standard terms);

[0102] Build a solution verification sandbox: simulate strategy implementation in the digital twin system (with an accuracy of 92%) and calculate the root mean square error between expected and actual effects (RMSE < 0.15).

[0103] According to one embodiment of the present application, step S6 of the teaching management method driven by learning performance evaluation based on capability map construction specifically includes:

[0104] S61: Construct a knowledge transfer map. Data preparation: Collect interdisciplinary project data (such as STEM course completion records) and annotate the ability transfer path (such as "programming ability → physical modeling")

[0105] Graph construction: nodes represent capability items (a total of 287 nodes), edge weight calculation: Jaccard similarity (number of co-occurring items / total number of items), application of FastRP algorithm for graph embedding (output 128-dimensional vector)

[0106] Dynamic update mechanism: The graph is updated incrementally every month (calculation of new edge weights takes less than 15 minutes), and the node influence score is calculated using the PageRank algorithm;

[0107] S62: Mining implicit ability associations and implementing association rule mining: Using the FP-Growth algorithm to discover frequent item sets (support ≥ 0.3, confidence ≥ 0.7). Example rule: "Circuit design ability ↑ → Mathematical modeling ability ↑ (improvement probability 62%)"

[0108] Construct a causal inference model: Use the PC algorithm to construct a structural equation model (SEM) and calculate the direct influence coefficient between abilities (e.g., β = 0.42 indicates the degree of direct influence).

[0109] Develop a migration path recommender: Input current capability shortcomings and output a solution containing three optimal migration paths (e.g., "improve data processing capabilities through chemical experiments");

[0110] S63: Design a collaborative training program and construct a discipline synergy matrix: calibrate the overlap between disciplines (e.g., the overlap between STEM disciplines reaches 58%) and calculate the synergy gain coefficient (γ = 1.2 means the joint training effect is improved by 20%).

[0111] Develop a course scheduling optimizer: input class hour constraints (total class hours per week ≤ 35), and output a multidisciplinary course combination (meeting the cross-coverage requirements of capabilities)

[0112] Implementation effect prediction: Monte Carlo simulation is used to predict the improvement in capability (with a 95% confidence interval) and generate risk warnings (e.g., "Insufficient physical experiment equipment may lead to a 30% delay in progress").

[0113] According to one embodiment of the present application, step S7 of the teaching management method driven by learning performance evaluation based on capability map construction specifically includes:

[0114] S71: Build a multi-dimensional user profile. Data collection: Academic data: 100+ ability indicators from the past six months; Behavioral data: average daily study time, resource access patterns (e.g., video viewing completion rate); Physiological data: eye movement heat map, heart rate variability (HRV);

[0115] Feature Engineering: Calculate learning style dimensions (visual / auditory / practical) and extract time management features (distribution of focused time periods, utilization of fragmented time);

[0116] Portrait storage: Use Redis to store real-time portraits (update cycle = 5 minutes) and build a historical portrait version library (retain snapshots of the last 30 days);

[0117] S72: Implement path planning algorithms and build a resource knowledge graph: Nodes: teaching resources (videos / question banks / experimental modules); Edges: applicable relationships (must meet pre-conditioning requirements)

[0118] Applying the ant colony optimization algorithm:

[0119] Pheromone update rule: τ ij =(1-ρ)τ ij +Δτ ij ; ρ = 0.5 is the pheromone volatility coefficient, and Δτ is positively correlated with resource availability;

[0120] Heuristic function design:

[0121] d ij is the resource difficulty difference, w k Knowledge coherence weight

[0122] Path pruning strategy: remove repeated coverage of competence items to ensure that the daily learning load is less than or equal to the cognitive load threshold (calculated by Kirschner formula);

[0123] S73: Generate dynamic adjustment plans and implement online learning analysis: monitor learning progress in real time (such as detecting video freeze points) and calculate the knowledge mastery decay rate (based on the Ebbinghaus forgetting curve)

[0124] Build an adaptive adjustment engine: When the learning efficiency drops for two consecutive days, it triggers path replanning and uses the simulated annealing algorithm to optimize the new path (termination temperature T = 0.01)

[0125] Design a visual navigation interface: display a capability development roadmap (with timeline predictions) and provide alternative path choices (such as "theoretical reinforcement mode" or "practical acceleration mode").

[0126] According to one embodiment of the present application, step S8 of the teaching management method driven by learning performance evaluation based on capability map construction specifically includes:

[0127] S81: Construct multimodal evaluation indicators, process indicator collection: classroom participation: voice sentiment analysis (positive / neutral / negative ratio), resource interaction depth: number of video pauses, note annotation density;

[0128] Calculation of outcome indicators: Knowledge mastery: IRT item response theory calculation of theta value; skill proficiency: code complexity score given by the automatic scoring system; indicator standardization: using Min-Max normalization processing (interval [0,1]);

[0129] S82: Implement a comprehensive evaluation model and build an evidence theory framework: Identify conflicting evidence (e.g., test scores and behavioral performance are inconsistent) and apply Dempster's combination rules to integrate them (initiate manual arbitration when the conflict degree is > 0.3)

[0130] Computational Ability Development Index: Longitudinal comparison: Comparison with historical baseline (calculation of Cohen's d effect size) Horizontal comparison: Comparison with peers (percentile ranking)

[0131] Generate radar chart analysis report: mark the advantage capability area (Z>1.5σ), mark the area to be improved (Z<-1σ), and display the capability association network (highlighting the critical path);

[0132] S83: Conduct teaching effectiveness diagnosis and construct an attribution model: Use Shapley values ​​to decompose the contribution of each factor (e.g., "teacher guidance contributes 32%)" and identify nonlinear relationships (e.g., "moderate pressure improves learning efficiency to a point where it then declines").

[0133] Develop root cause analysis tools: Implement decision tree feature importance ranking (e.g., "frequency of classroom interaction is a key factor") and generate actionable recommendations (e.g., "increase the frequency of group discussions to three times a week");

[0134] Build a continuous improvement dashboard: display the teaching effectiveness index trend (90-day moving average) and mark classes / subjects that require key intervention.

[0135] According to one embodiment of the present application, step S9 of the teaching management method driven by learning performance evaluation based on capability map construction specifically includes:

[0136] S91: Build a resource evaluation system and design multi-dimensional evaluation indicators: Content quality: expert review (5-point scale) + student rating (1-5 stars); Matching degree: cosine similarity based on ability map; Teaching effectiveness: use the IRT model to calculate the discrimination parameter (a>0.65 is excellent)

[0137] Implement online evaluation: Automatically collect feedback (length of stay, number of repeat visits) after each resource use and calculate the resource health index: H = 0.5Q + 0.3M + 0.2E; where Q is the quality score, M is the matching degree, and E is the teaching effect;

[0138] Construct resource life cycle model: Introduce product life cycle theory to divide the stages (introduction period / growth period / decline period) and predict resource failure time (Weibull distribution parameter estimation);

[0139] S92: Implement intelligent recommendation strategies and build hybrid recommendation models: collaborative filtering (user-resource matrix decomposition, latent vector dimension = 64), content recommendation (BERT text similarity calculation);

[0140] Implement context-aware recommendations: consider time factors (recommending memory resources in the morning) and adapt to device types (prioritizing short videos on mobile devices);

[0141] Develop a recommendation explanation module: generate top-N recommendation reasons (e.g., "Because your spatial reasoning ability is weak, the 3D geometry module is recommended") and display similar user selection paths (collaborative filtering explanation);

[0142] S93: Build a resource evolution mechanism and implement an A / B testing framework: run the old and new versions of resources in parallel (sample size ≥ 300 / group), and use t-tests to compare the effects (significance level α = 0.05);

[0143] Develop an automatic iteration system: automatically replace the old version when the CTR of the new version of the resource increases by ≥15%, and build a genetic algorithm to optimize the content structure (crossover rate = 0.8, mutation rate = 0.1);

[0144] Design a resource elimination pipeline: resources that have not been accessed for 30 consecutive days will be downgraded, and inefficient resources will be marked and put into the optimization pool (for teachers' secondary creation).

[0145] The above-described embodiments merely represent several implementations of the present invention. While the descriptions are relatively specific and detailed, they are not to be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the spirit of the present invention, and such modifications and improvements fall within the scope of the present invention. Therefore, the scope of the present invention shall be determined by the claims.

Claims

1. A teaching management method driven by learning performance evaluation based on capability map construction, characterized by: The steps include: S1: Steps for constructing a multidimensional capability map, including determining capability dimensions and indicator systems, establishing capability association models, developing map visualization tools, and developing capability assessment tools; S2: Steps for implementing a multimodal data acquisition system, including deploying intelligent sensing devices, implementing data cleaning and fusion, and building a data storage architecture; S3: Steps for developing a dynamic capability value calculation model, including establishing a differential equation model, implementing capability prediction, and generating a capability development report; S4: Steps for building a real-time capability warning system, including building a real-time data stream processing pipeline, designing a multi-level warning rule engine, and implementing a hierarchical response mechanism; S5: Steps for generating adaptive teaching strategies, including building a teaching action feature library, implementing reinforcement learning strategy optimization, and generating explainable teaching plans; S6: Steps for conducting cross-disciplinary competence correlation analysis, including constructing a knowledge transfer map, mining implicit competence correlations, and designing collaborative training programs; S7: Steps for generating personalized learning paths, including building multi-dimensional user profiles, implementing path planning algorithms, and generating dynamic adjustment plans; S8: Steps for conducting dynamic evaluation of teaching effectiveness, including constructing multimodal evaluation indicators, implementing a comprehensive evaluation model, and conducting teaching effectiveness diagnosis; S9: Steps for dynamic optimization of teaching resources, including building a resource evaluation system, implementing intelligent recommendation strategies, and building a resource evolution mechanism.

2. The teaching management method driven by learning performance evaluation based on capability map construction according to claim 1 is characterized in that: The step S1 specifically includes: S11: Determine the competency dimensions and indicator system. Organize an interdisciplinary expert group, including educational psychologists, subject leaders, and data scientists, and use three rounds of Delphi to determine the core competency framework: the first round collects open-ended opinions and forms 128 potential competency items; the second round uses a 5-point Likert scale to assess importance and screens out 42 key competencies; the third round verifies the indicator discrimination and calculates the Cronbach's alpha coefficient for each indicator, which is required to be greater than 0.7; Construct a three-level indicator tree: First-level dimension: knowledge reserve K, skill level S, cognitive literacy C, emotional attitude A; Second-level indicator: knowledge dimension includes concept mastery K1, knowledge transfer ability K2; Third-level indicator: detailed application of the fundamental theorem of calculus K1-3, interdisciplinary analogical reasoning C2-1; Weight calculation: Use the hierarchical analysis method to construct a judgment matrix and calculate the weight of each level. The weight of the knowledge dimension is 0.35, of which the weight of K1 is 0.4, the weight of K2 is 0.3, and the weight of K3 is 0.

3. S12: Establish a capability association model and construct a capability migration map: nodes represent capability items, and edge weights are calculated using historical data; Develop graph visualization tools: use Neo4j graph database to store relational data; implement interactive query functions; Dynamic update mechanism: collect 2000+ pieces of capability development data each semester; use PageRank algorithm to iteratively update node importance scores; S13: Develop competency assessment tools and design multimodal assessment scales: Knowledge dimension: Adaptive testing is used to dynamically generate 30-50 questions per session; Skill dimension: Project analysis is performed; Emotion dimension: Semantic analysis technology is used to process learning logs; Build a calibration dataset: Collect at least 1,000 student sample data and annotate the ability development trajectory; calculate the reliability and validity of the assessment tool, and deploy an automated scoring engine: Use the BERT model to perform semantic understanding on subjective questions and generate instant feedback reports.

3. The teaching management method driven by learning performance evaluation based on capability map construction according to claim 1 is characterized in that: Step S2 specifically includes: S21: Deploy intelligent sensing devices and build an IoT data collection network: Deploy eight 4K cameras in classrooms to capture facial expressions, install millimeter-wave radars to monitor student posture changes, and equip smart bracelets to record physiological indicators. Configure the learning behavior tracking module: the LMS system records resource access paths, and the eye tracker tracks gaze heat maps; Data synchronization mechanism: Use timestamps to align multi-source data, build a data lineage map, and record the collection device and timestamp of each data point; S22: Implement data cleaning and fusion, noise elimination processing: apply wavelet threshold denoising to camera data and perform Kalman filtering on physiological signals; Feature engineering: extracting temporal features and calculating spatial features; Data fusion: Use the federated learning framework to aggregate data across schools and apply DS evidence theory to fuse multi-source data; S23: Build a data storage architecture and design a tiered storage solution: Raw data layer: HDFS stores raw video / sensor data; Feature data layer: MongoDB stores time series features; Knowledge data layer: Neo4j stores capability association graphs; Implement data lifecycle management: hot data is stored in SSD clusters, warm data is migrated to Blu-ray storage, and cold data is transferred to tape libraries; Security protection mechanism: Apply homomorphic encryption technology to protect sensitive data and implement blockchain-based data audit logs.

4. The teaching management method driven by learning performance evaluation based on capability map construction according to claim 1 is characterized in that: Step S3 specifically includes: S31: Establish a differential equation model and construct the ability development dynamics equation: in: A(t): ability value at time t, 0≤A≤1; K: Ability limit; r: individual learning efficiency coefficient; t0: Ability inflection point time, when the moment of greatest strength; λ: knowledge internalization coefficient; I(τ): instructional input intensity function; Parameter calibration uses the least squares method to fit historical data. Boundary processing forces normalization to the K value when A(t)>K. The model integrates biological growth curves and cognitive load theory. When the teaching input intensity I(τ) exceeds a critical value, the system automatically switches to "deep learning mode" to increase the supply of case analysis resources. Capacity values ​​are calculated in batches at 2:00 AM daily, with an error rate within ±3%. Develop a numerical solver: Use the fourth-order Runge-Kutta method and set convergence conditions. Build an anomaly detection mechanism: When the calculated value deviates from the measured value by more than 15%, trigger the calibration process and automatically adjust the λ parameter. S32: Implement capability prediction and train the LSTM prediction model: input layer: capability value sequence of the past 30 days, hidden layer: 256 LSTM units, output layer: capability prediction value of the next 7 days; Constructing probabilistic prediction intervals: Using quantile regression to generate 95% confidence intervals, triggering an alert when the lower bound of the prediction falls below a threshold; Implement dynamic calibration: fine-tune model parameters using new data every week and calculate the forecast error MAPE; S33: Generate capability development report and visualize it: draw capability development curve and generate capability short board radar chart; Form intervention recommendations: match the knowledge gap database and calculate resource matching; automate report generation, use the template engine to dynamically fill in content, and output PDF / HTML dual-format reports.

5. The teaching management method driven by learning performance evaluation based on capability map construction according to claim 1 is characterized in that: The step S4 specifically includes: S41: Deploy a distributed stream processing framework: Data source access: Convert heterogeneous data from cameras, sensors, and LMS systems into JSON format. Sharding strategy: Hash sharding by student ID to ensure data continuity for the same student. Window calculation: Set a sliding window to calculate the volatility of the statistical capability indicator. The calculation formula is: When volatility > 0.3, an abnormal flag is triggered; Implement feature engineering: Calculate time-domain features using rolling averages and exponentially weighted moving averages; extract frequency-domain features using FFT transforms to analyze the low-frequency energy ratio of heart rate variability; build a status monitoring dashboard that displays a real-time heat map of the ability distribution across campuses, identifies abnormal student locations, and automatically generates a vector map of ability development speed; S42: Establish a dynamic threshold system: Baseline threshold: Calculate a moving average based on historical data; Adaptive threshold: Automatically adjust the threshold when the class average ability value changes by ±5%; Personalized threshold: Set differentiated parameters for special student groups; The rule logic tree is constructed as follows: A[ability value < 0.2] --> B{lasts for 3 days?}; B-->|is|C[red alert]; B-->|No|D[Orange Alert]; A-->E{Is there any escape behavior?}; E-->|is|F[red alert]; Develop rule verification, using decision trees to verify rule coverage; implement conflict resolution, prioritize the highest confidence rule when multiple rules are triggered simultaneously, and perform rule stress testing monthly; S43: Construct an early warning response knowledge base: a hierarchical intervention plan library: including 6 basic plans and 32 extended plans, a learning resource index table: establish an inverted index by capability item; deploy an intelligent decision-making engine, use Bayesian networks to calculate the posterior probability of intervention measures, and implement cost-benefit analysis: calculate the ratio of time cost to expected benefits; design a closed-loop feedback system, record the execution of each intervention, and calculate the strategy effectiveness index.

6. The teaching management method driven by learning performance evaluation based on capability map construction according to claim 1 is characterized in that: The step S5 specifically includes: S51: Build a teaching action feature library and collect teaching intervention samples: record more than 100,000 teaching cases from the past three years; annotate key features: teaching methods, resource types, and duration distribution; vectorize features, use TF-IDF to extract text features, and perform CNN feature extraction on video resources; construct a similarity matrix, calculate cosine similarity, and generate a teaching action clustering map; S52: Implement reinforcement learning strategy optimization, design state space: current ability map state (, learning environment characteristics; define action space, basic actions, combined actions; Construct the reward function: R(s,a)=0.6ΔA+0.3ΔE+0.1ΔT; ΔA: ability improvement rate; ΔE: Change in learning efficiency, the increment of knowledge points per unit time; ΔT: reduction in resource consumption time; Training the DQN model: using the Double DQN algorithm to eliminate overestimation, experience replay pool capacity = 1e6, prioritizing samples with high TD error, target network update frequency = 100 steps; S53: Generate explainable teaching plans, implement strategy counterfactual analysis, compare the differences between the current strategy and the historical optimal plan, and mark key decision factors; develop a natural language generator to generate strategy descriptions based on the BERT model and output guidance suggestions that conform to educational principles; build a solution verification sandbox, simulate strategy implementation in the digital twin system, and calculate the root mean square error between expected and actual effects.

7. The teaching management method driven by learning performance evaluation based on capability map construction according to claim 1 is characterized in that: The step S6 specifically includes: S61: Construct a knowledge transfer graph, including data preparation, collecting interdisciplinary project data, and labeling capability transfer paths; construct a graph, represent capability items with nodes, calculate edge weights, perform Jaccard similarity, and apply the FastRP algorithm for graph embedding; implement a dynamic update mechanism, incrementally update the graph monthly, and calculate node influence scores using the PageRank algorithm; S62: Uncover implicit ability associations, implement association rule mining, and use the FP-Growth algorithm to discover frequent itemsets; construct a causal inference model, apply the PC algorithm to construct a structural equation model, and calculate the direct influence coefficients between abilities; develop a migration path recommender that inputs current ability shortcomings and outputs a solution containing three optimal migration paths; S63: Design collaborative training programs, construct a disciplinary collaboration matrix, calibrate the degree of overlap in interdisciplinary capabilities, and calculate the collaborative gain coefficient; develop a course scheduling optimizer, input class hour constraints, and output a multidisciplinary course combination; implement effect prediction, use Monte Carlo simulation to predict the extent of capability improvement, and generate risk warnings.

8. The teaching management method driven by learning performance evaluation based on capability map construction according to claim 1 is characterized in that: The step S7 specifically includes: S71: Build a multi-dimensional user profile. Data collection includes: academic data (over 100 ability indicators from the past six months); behavioral data (average daily study time and resource access patterns); physiological data (eye movement heatmaps and heart rate variability); feature engineering (calculating learning style dimensions and extracting time management features); and profile storage (using Redis to store real-time profiles and build a historical profile version library). S72: Implement path planning algorithm and build resource knowledge graph: nodes include teaching resources; edges include applicable relationships, and apply ant colony optimization algorithm: Pheromone update rule: τ ij =(1-ρ)τ ij +Δτ ij ; ρ = 0.5 is the pheromone volatility coefficient, and Δτ is positively correlated with resource availability; Heuristic function design: d ij is the resource difficulty difference, w k Knowledge coherence weight Path pruning strategy: remove repeated competency items to ensure that daily learning load is less than or equal to the cognitive load threshold; S73: Generate dynamic adjustment plans, implement online learning analysis, monitor learning progress in real time, and calculate the knowledge mastery decay rate; build an adaptive adjustment engine that triggers path replanning when it detects a decrease in learning efficiency for two consecutive days, and uses a simulated annealing algorithm to optimize the new path; design a visual navigation interface to display the capability development roadmap and provide alternative path selection.

9. The teaching management method driven by learning performance evaluation based on capability map construction according to claim 1 is characterized in that: The step S8 specifically includes: S81: Construct multimodal evaluation indicators, process indicator collection, classroom participation: voice sentiment analysis, resource interaction depth: number of video pauses, note annotation density; result indicator calculation: knowledge mastery: IRT item response theory calculation of theta value, skill proficiency: code complexity score given by the automatic scoring system; indicator standardization: use Min-Max normalization processing; S82: Implement a comprehensive evaluation model and build an evidence-based theoretical framework: Identify conflicting evidence and integrate it using Dempster's combination rule; calculate the capability development index; conduct longitudinal comparisons with historical baselines and horizontal comparisons with peer groups; generate radar chart analysis reports, annotate areas of superior capabilities, mark areas for improvement, and display the capability association network; S83: Implement teaching effectiveness diagnosis, build attribution models, use Shapley values ​​to decompose the contribution of each factor, and identify nonlinear relationships; develop root cause analysis tools, implement decision tree feature importance ranking, and generate actionable suggestions; build a continuous improvement dashboard to display the teaching effectiveness index trend and mark classes / subjects that require key intervention.

10. The teaching management method driven by learning performance evaluation based on capability map construction according to claim 1 is characterized in that: The step S9 specifically includes: S91: Construct a resource evaluation system and design multi-dimensional evaluation indicators for content quality, including expert review and student ratings; the matching degree is based on the cosine similarity of the ability map; the teaching effect uses the IRT model to calculate the discrimination parameter; Implement online evaluation, automatically collect feedback after each resource use, and calculate the resource health index: H = 0.5Q + 0.3M + 0.2E; where Q is the quality score, M is the matching degree, and E is the teaching effect; build a resource life cycle model, introduce the product life cycle theory to divide the stages, and predict the time of resource expiration; S92: Implement intelligent recommendation strategies and build a hybrid recommendation model for collaborative filtering and content recommendation. Implement context-aware recommendations, taking time into account, adapting to device types, generating top-N recommendation reasons, and displaying similar user selection paths. S93: Build a resource evolution mechanism, implement an A / B testing framework, run old and new versions of resources in parallel, and use t-tests to compare the differences in effects; develop an automatic iteration system to automatically replace the old version when the CTR of the new version of the resource increases by ≥15%, and build a genetic algorithm to optimize the content structure; design a resource elimination pipeline to downgrade resources that have not been accessed for 30 consecutive days, and mark inefficient resources into the optimization pool.

Images