Teaching quality evaluation and diagnosis method and system based on dynamic credibility weighting and knowledge graph

Abstract

The application discloses a teaching quality evaluation diagnosis method and system based on dynamic credibility weighting and a knowledge graph, comprising the following steps: multi-source heterogeneous data acquisition and standardized preprocessing: collecting teaching process data, subjective evaluation data, teaching background data and teacher development data to form a standardized data set; credibility dynamic weighting and fusion evaluation: calculating the basic credibility of each data source or evaluator to generate a credibility fusion evaluation score with a confidence interval or credibility level; teaching quality dynamic portrait construction and real-time early warning: connecting the credibility fusion evaluation scores in time sequence to construct a dynamic teaching portrait and monitoring the real-time data inflow based on preset business rules; and individualized diagnosis report intelligent generation: obtaining the significant advantage dimension and the dimension to be improved of a teaching unit, associating a teaching improvement knowledge graph and generating a structured individualized diagnosis report.

Description

Technical Field

[0001] This invention relates to the fields of educational informatization and data mining technology, and in particular to a teaching quality assessment and diagnosis method and system based on dynamic credibility weighting and knowledge graphs. Background Technology

[0002] The teaching quality assessment and diagnosis method based on dynamic credibility weighting and knowledge graphs uses dynamic credibility weighting as its core mechanism. By comprehensively considering dynamic factors, it assigns differentiated weights to indicators in each stage of teaching to achieve adaptive optimization of the assessment model. Simultaneously, it relies on knowledge graphs to construct a network of connections between teaching elements, structurally integrating multi-source heterogeneous data to form a multi-dimensional teaching knowledge graph. Then, through graph reasoning and path analysis, it achieves precise location and causal tracing of teaching problems. Its aim is to construct a scientific, objective, and dynamic teaching quality assessment system. This system overcomes the assessment distortion caused by fixed weights in traditional methods and breaks through the limitations of single-dimensional data silos, achieving a transformation from an "experience-driven" to a "data + knowledge dual-driven" assessment model. This improves the accuracy and predictability of teaching quality diagnosis, providing scientific decision support for improving teachers' teaching abilities, optimizing curriculum design, and allocating educational resources. It promotes a paradigm shift in teaching assessment from outcome evaluation to process diagnosis, from single evaluation to multi-dimensional diagnosis, and from static description to dynamic prediction, thereby contributing to the continuous improvement of educational quality and the intelligent development of the educational ecosystem.

[0003] Existing teaching evaluation systems, while attempting to integrate data, mostly remain at the level of simple data collection and report display. They fail to build a closed-loop system from a technical perspective that can automatically handle data inconsistencies, dynamically weight and evaluate evidence, identify teaching risks in real time, and generate intelligent diagnoses. Therefore, an innovative technical solution is urgently needed to address these issues. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of the prior art by proposing a teaching quality assessment and diagnosis method and system based on dynamic credibility weighting and knowledge graph.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: The teaching quality assessment and diagnostic method based on dynamic credibility weighting and knowledge graphs includes the following steps: Multi-source heterogeneous data acquisition and standardized preprocessing: Through API interface, database synchronization and log parsing technology, teaching process data, subjective evaluation data, teaching background data and teacher development data are automatically collected. The collected data are cleaned, aligned and standardized. Natural language processing technology is used to quantify text comments, extract feature vectors and form a standardized dataset with teacher-course-teaching week as the basic spatiotemporal unit. Credibility-based dynamic weighted fusion assessment: Based on the standardized dataset, the basic credibility of each data source or evaluator is calculated. By analyzing the consistency, correlation, or conflict between the assessment indicators reflected by different data sources within the same teaching unit, the cross-validation results of the multi-source data of the current teaching unit are obtained. The basic credibility weights are dynamically adjusted according to the cross-validation results. The multi-dimensional assessment indicators are weighted and fused according to the adjusted dynamic weights to generate a credibility fusion assessment score with confidence intervals or credibility levels. Construction and real-time early warning of dynamic teaching quality profile: The trusted fusion evaluation scores are connected in time sequence to construct a dynamic teaching profile. Based on preset business rules, the real-time incoming data is monitored. When threshold early warning, trend early warning, comparison early warning or conflict early warning is triggered, an early warning event is generated and pushed to the specified object. Intelligent generation of personalized diagnostic reports: Based on predefined identification rules, the credible fusion evaluation score is analyzed to obtain the significant advantages and dimensions to be improved of the teaching unit. The teaching improvement knowledge graph is linked, and matching improvement cases, teaching strategies and related resources are retrieved according to the identified dimensions to generate a structured personalized diagnostic report.

[0006] The above technical solution further includes: Furthermore, the teaching process data includes student video viewing time, assignment submission time, forum posts, chapter test scores, and login frequency behavior logs obtained from the online learning platform; the subjective evaluation data includes quantitative scores and text comments from students, supervisors, and peers obtained from the teaching evaluation system, as well as class observation records obtained from the class observation system; the teaching background data includes course attributes, student grades, class sizes, and academic performance data obtained from the academic affairs system; and the teacher development data includes teaching training records, teaching reform projects, and awards obtained from teacher files.

[0007] Furthermore, calculating the basic credibility of each data source or evaluator includes: For supervisory and peer evaluations, the basic credibility is calculated based on the correlation coefficient of long-term consistency between their historical evaluations and the comprehensive benchmark. For student evaluations of teaching, the basic credibility is calculated by comprehensively considering the consistency of all student evaluations, the length of the evaluation text, and the degree of dispersion with other student evaluations. For teaching process data, the basic reliability is calculated based on the completeness and stability of the data.

[0008] Furthermore, dynamically adjusting the basic confidence weights based on the cross-validation results specifically includes: When the performance of multiple data sources within the same teaching unit corroborates each other, the temporary weight of the corresponding data source is increased. When the evaluation results of the data source conflict significantly with the conclusions of other data sources, a verification mechanism is triggered and the dynamic weight of the data source is reduced. The cross-validation covers the correlation analysis between teaching process data, subjective evaluation data, teaching background data, and teacher development data.

[0009] Furthermore, the preset business rules include threshold warning rules, trend warning rules, comparison warning rules, and conflict warning rules: The threshold warning rule means that a warning will be issued if a student's online learning time, timely homework submission rate, and frequency of classroom interaction exceed a preset fluctuation range for several consecutive weeks. The trend warning rule is expressed as identifying a significant downward trend in indicators such as classroom interaction index and student engagement through linear regression analysis; The comparison warning rule means that a warning is triggered when the score of a certain dimension is consistently lower than the preset range of the average level of similar courses in the same college. The conflict warning rule means that when there are high contradictions in multi-source data within the same teaching unit, an abnormality in the evaluation process will be indicated and a warning will be generated.

[0010] Furthermore, the teaching improvement knowledge graph is a structured knowledge base that stores the relationships between teaching problems, causes, countermeasures, successful cases, and related resources. Based on the problem dimensions identified by the trusted fusion evaluation score, the report generation engine retrieves matching countermeasures and cases from the knowledge graph and automatically organizes and generates a personalized diagnostic report containing specific improvement suggestions, reference cases, and resource links.

[0011] A teaching quality assessment and diagnostic system based on dynamic credibility weighting and knowledge graph includes: Multi-source data acquisition and preprocessing module: Collects and standardizes multi-source heterogeneous data from four types of data sources: teaching process, subjective evaluation, teaching background and teacher development, and quantifies text comments through natural language processing technology; Credibility Dynamic Weighted Fusion Evaluation Module: Dynamically calculates credibility weights for each data source or evaluator, adjusts weights based on cross-validation of multi-source data, and performs weighted fusion of multi-dimensional evaluation indicators to generate a credibility fusion evaluation score with confidence intervals; Teaching quality dynamic early warning module: Constructs a time-series dynamic teaching profile, monitors real-time data based on preset rules, and automatically generates and pushes early warning events when early warning conditions are triggered; Personalized diagnostic report generation module: Based on the credible fusion assessment score, it identifies teaching strengths and areas for improvement, and automatically generates a structured diagnostic report by combining the teaching improvement knowledge graph; Visual interactive platform: Displays dynamic profiles, early warning information, assessment results and diagnostic reports, and provides a human-computer interaction interface.

[0012] Furthermore, the credibility dynamic weighted fusion evaluation module includes a basic credibility calculation unit, a dynamic weight adjustment unit, and a fusion evaluation score calculation unit: The basic credibility calculation unit calculates the basic credibility weights of various data sources based on the evaluator's historical consistency, the student's evaluation discrimination and seriousness, and the completeness of teaching process data. The dynamic weight adjustment unit adjusts the weight of each data source in the current teaching unit in real time based on the consistency, correlation or conflict between data through a multi-source data cross-validation mechanism. The fusion evaluation score calculation unit performs weighted fusion of standardized multi-dimensional evaluation indicators based on dynamic weights, and outputs a credible fusion evaluation score with a credibility level.

[0013] Furthermore, the dynamic early warning module for teaching quality includes a profile building unit, a rule engine unit, and an early warning generation and push unit: The portrait construction unit connects the multi-dimensional credible integrated evaluation scores of each teaching unit in chronological order to form a visually representable dynamic teaching status trend chart. The rule engine unit embeds four types of early warning rules: threshold, trend, comparison, and conflict, which automatically match and judge the real-time incoming teaching data stream. The warning generation and push unit automatically generates warning events when a rule is triggered, and pushes the warning information to teaching administrators and teachers' terminals in real time through a message queue or interface.

[0014] Furthermore, the personalized diagnostic report generation module includes a problem identification unit, a knowledge graph query unit, and a report synthesis unit: The problem identification unit analyzes the scores and trends of the credible fusion evaluation score in each dimension, and identifies the dimensions with significant advantages and dimensions that need improvement. The knowledge graph query unit is linked to built-in or external teaching improvement knowledge graphs, and retrieves related cause analyses, improvement strategies, success cases and teaching resources based on the identified problems. The report synthesis unit automatically organizes the search results into a structured text report that includes problem diagnosis, specific suggestions, reference cases and resource links, and supports output in document or visualization form.

[0015] The present invention has the following beneficial effects: In this invention, by generating feature vectors of analysis parameters that match the service type and dynamic baselines, the system can automatically adjust the analysis parameters and judgment baselines to adapt to normal behavior patterns under different service types and different spatiotemporal modes, greatly reducing the workload of manual configuration and maintenance. Attached Figure Description

[0016] Figure 1 This is a flowchart of the teaching quality assessment and diagnosis method based on dynamic credibility weighting and knowledge graph proposed in this invention; Figure 2 This is a system block diagram of the teaching quality assessment and diagnosis system based on dynamic credibility weighting and knowledge graph proposed in this invention. Detailed Implementation

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

[0018] Please see Figures 1-2 As shown, this invention is a teaching quality assessment and diagnosis method based on dynamic credibility weighting and knowledge graphs, comprising the following steps: Multi-source heterogeneous data acquisition and standardized preprocessing: Through API interface, database synchronization and log parsing technology, teaching process data, subjective evaluation data, teaching background data and teacher development data are automatically collected. The collected data are cleaned, aligned and standardized. Natural language processing technology is used to quantify text comments, extract feature vectors and form a standardized dataset with teacher-course-teaching week as the basic spatiotemporal unit. Credibility-based dynamic weighted fusion assessment: Based on the standardized dataset, the basic credibility of each data source or evaluator is calculated. By analyzing the consistency, correlation, or conflict between the assessment indicators reflected by different data sources within the same teaching unit, the cross-validation results of the multi-source data of the current teaching unit are obtained. The basic credibility weights are dynamically adjusted according to the cross-validation results. The multi-dimensional assessment indicators are weighted and fused according to the adjusted dynamic weights to generate a credibility fusion assessment score with confidence intervals or credibility levels. Construction and real-time early warning of dynamic teaching quality profile: The trusted fusion evaluation scores are connected in time sequence to construct a dynamic teaching profile. Based on preset business rules, the real-time incoming data is monitored. When threshold early warning, trend early warning, comparison early warning or conflict early warning is triggered, an early warning event is generated and pushed to the specified object. Intelligent generation of personalized diagnostic reports: Based on predefined identification rules, the credible fusion evaluation score is analyzed to obtain the significant advantages and dimensions to be improved of the teaching unit. The teaching improvement knowledge graph is linked, and matching improvement cases, teaching strategies and related resources are retrieved according to the identified dimensions to generate a structured personalized diagnostic report.

[0019] In one embodiment, the teaching process data includes student video viewing time, assignment submission time, forum posts, chapter test scores, and login frequency behavior logs obtained from the online learning platform; the subjective evaluation data includes quantitative scores and text comments from students, supervisors, and peers obtained from the teaching evaluation system, as well as class observation records obtained from the class observation system; the teaching background data includes course attributes, student grade, class size, and academic performance data obtained from the academic affairs system; and the teacher development data includes teaching training records, teaching reform projects, and awards obtained from teacher files.

[0020] In one embodiment, calculating the basic credibility of each data source or evaluator includes: For supervisory and peer evaluations, the basic credibility is calculated based on the correlation coefficient of long-term consistency between their historical evaluations and the comprehensive benchmark. It should be noted that the specific analytical process for calculating the reliability of the data sources for supervision and peer review is as follows: Define the comprehensive benchmark: The comprehensive benchmark is the student achievement gain of the evaluated course in subsequent teaching stages, or the long-term impact score of the course based on graduate feedback surveys many years later. Collect historical evaluation data: Extract the historical evaluation scores of all courses by the supervisor or peer within a preset time window (such as the past 2 years) to form a historical evaluation sequence; Calculate the long-term consistency correlation coefficient: Perform correlation analysis between the historical evaluation sequence and the corresponding comprehensive benchmark sequence of the course, and calculate the Pearson correlation coefficient as a consistency indicator. : , The value ranges from [-1, 1]. The closer the value is to 1, the higher the positive consistency between the evaluator's historical evaluation and the course's long-term performance. For their respective standard deviations, Extract the n valid course evaluations completed by the evaluator within a preset historical period. The standardized score representing the i-th evaluation is used to simultaneously obtain the comprehensive benchmark value for the corresponding n courses. For the student performance gain or long-term feedback rating of the i-th course, Let n be the arithmetic mean of n valid course evaluations. The arithmetic mean of the combined baseline values ​​of n courses; Calculating the baseline credibility weight: Introducing the number of evaluations (NN) as a confidence factor, the baseline credibility weight of supervision or peer review is calculated using the following formula. : ,in, For adjustment coefficients, The minimum number of valid evaluations is set as the threshold, and the consistency correlation coefficient is... The higher and the more evaluations The more sufficient the data, the higher the calculated base credibility weight. The higher; For student evaluations of teaching, the basic credibility is calculated by comprehensively considering the consistency of all student evaluations, the length of the evaluation text, and the degree of dispersion with other student evaluations. It should be noted that the specific analytical process for calculating the reliability of the data sources for supervision and peer review is as follows: Quantitative evaluation consistency index: Calculate the standard deviation of all teaching evaluation scores given to this student during the semester. A lower standard deviation indicates higher evaluation consistency; therefore, a consistency factor is set. ; Quantitative evaluation of seriousness indicators: Based on the total character length L, the mean absolute value of emotional intensity E, and the number of keywords covered K in the student's submitted text comments, a seriousness factor is calculated through weighting. ,in, For normalization function, The weighting coefficients and ; Quantitative evaluation of dispersion index: Calculate the absolute difference D between the student's evaluation score for the current course and the average evaluation score of all students in the class for the course, and normalize it into a dispersion factor based on the dispersion of the whole class. ; Comprehensive calculation of basic credibility weight: The above three factors are combined to obtain the basic credibility weight of the student's single evaluation. , The fusion weighting coefficients are adjustable, and ; For teaching process data, the basic reliability is calculated based on the completeness and stability of the data.

[0021] It should be noted that the specific analytical process for calculating the reliability of the data sources for supervision and peer review is as follows: Calculate the data integrity index: For the current teaching unit, count the total number of data points that should be collected. The number of valid data points actually successfully collected Then the integrity ratio ; Calculate data stability indicators: Select key teaching process indicators (such as weekly login frequency and on-time homework submission rate) and calculate their coefficients of variation within adjacent periods of the teaching unit. (i.e., the ratio of standard deviation to mean), and converted into a stability factor. ; Comprehensive calculation of basic credibility weight: Based on the aforementioned integrity ratio and stability factor, the basic credibility weight of the data source for this teaching process is calculated through weighted average. ,in, This is an adjustment coefficient used to balance the importance of integrity, with a value range of 0 < <1.

[0022] In one embodiment, dynamically adjusting the basic confidence weights based on the cross-validation results specifically includes: When the performance of multiple data sources within the same teaching unit corroborates each other, the temporary weight of the corresponding data source is increased. When the evaluation results of the data source conflict significantly with the conclusions of other data sources, a verification mechanism is triggered and the dynamic weight of the data source is reduced. The cross-validation covers the correlation analysis between teaching process data, subjective evaluation data, teaching background data, and teacher development data.

[0023] It should be noted that the specific analytical process for dynamically adjusting the basic credibility weights is as follows: Definition and calculation of verification indicators: For any two data sources within the current teaching unit and The system calculates its evaluation dimensions for the same or related dimensions. Consistency score The score is obtained by comparing the correlation coefficient, trend similarity, or conclusion matching degree of the standardized data sequences (or evaluation conclusions) of the two data sources on this dimension; At the same time, calculate the conflict score. This score reflects the differences between the two data sources in terms of dimensions. The degree of contradiction in the above conclusions can be quantified by the dispersion of conclusions, the degree of trend deviation, or the logical contradiction. Determine the dynamic adjustment coefficient: Based on the aforementioned validation metrics, the system calculates a comprehensive validation factor for each data source within the current teaching unit. This factor aggregates the interaction validation results of this data source with all other relevant data sources across all evaluation dimensions; An example calculation formula is as follows: ,in, For data source In dimensions The basic weights or importance coefficients on the surface. , To adjust the parameters, The adjustment amount is the baseline. When multiple data sources corroborate each other, they exhibit correlation. The values ​​are generally high and The value is low, and the calculated value is low. A value greater than 1 (or a set baseline threshold) will result in an increased weight. When a data source significantly conflicts with other data sources, it manifests as correlation. The calculated value is significantly higher than expected. If the value is less than 1 (or the baseline threshold), a mechanism to reduce the weight will be triggered. Execution weights are dynamically adjusted: Data source Basic credibility weight Its integrated verification factor By combining these factors, we obtain their dynamic weights for the integrated assessment of the current teaching unit. ; The adjustment formula can be expressed as: , where the function It could be a simple multiplication operation, or a more complex normalization function, to ensure that the sum of all weights is a fixed value; This process covers the correlation analysis between teaching process data, subjective evaluation data, teaching background data, and teacher development data. That is, all other relevant data sources in the above calculations cover all types of these four data categories, ensuring that cross-validation is a comprehensive analysis across data categories. Output and Application: Calculated results This will be directly used for subsequent trusted fusion evaluation score calculation, enabling real-time and objective allocation of contributions from different data sources; The system records the basis for each weight adjustment as part of the interpretation of the confidence level of the evaluation results.

[0024] In one embodiment, the preset business rules include threshold warning rules, trend warning rules, comparison warning rules, and conflict warning rules: The threshold warning rule means that a warning will be issued if a student's online learning time, timely homework submission rate, and frequency of classroom interaction exceed a preset fluctuation range for several consecutive weeks. The trend warning rule is expressed as identifying significant downward trends in positive indicators such as classroom interaction index and student engagement, as well as significant downward trends in negative indicators such as absenteeism / truancy rate and frequency of mention of specific negative topics, through linear regression analysis. The comparison warning rule means that a warning is triggered when the score of a certain dimension is consistently lower than the preset range of the average level of similar courses in the same college. The conflict warning rule means that when there are high contradictions in multi-source data within the same teaching unit, an abnormality in the evaluation process will be indicated and a warning will be generated.

[0025] In one embodiment, the teaching improvement knowledge graph is a structured knowledge base that stores the relationships between teaching problems, causes, countermeasures, success cases, and related resources. The report generation engine, based on the problem dimensions identified by the trusted fusion evaluation score, retrieves matching countermeasures and cases from the knowledge graph and automatically organizes and generates a personalized diagnostic report containing specific improvement suggestions, reference cases, and resource links.

[0026] It should be noted that the specific analysis process for the intelligent generation of personalized diagnostic reports is as follows: The mapping from dimensions to be improved to knowledge graph queries is implemented, and the report generation engine receives a list of dimensions to be improved from the output of the dimension recognition unit. The system performs semantic matching or keyword mapping between each dimension to be improved and the predefined teaching problem node categories in the teaching improvement knowledge graph, thereby transforming the specific problems found in the assessment into queryable entry points for the knowledge graph. Based on the multi-hop retrieval and matching of association relationships, the engine takes the mapped teaching question node as the starting point and performs multi-hop retrieval in the teaching improvement knowledge graph; First, retrieve the associated potential causal nodes along the "problem-cause" edge; Secondly, retrieve targeted teaching strategies and improvement countermeasures nodes along the "cause-countermeasure" path; Finally, along the "Countermeasures-Case / Resources" edge, successful improvement cases and related training resource nodes are retrieved, forming a subgraph structure closely related to the dimension to be improved; Prioritizing and filtering countermeasures and case studies: The system sorts and filters multiple improvement countermeasures and successful cases based on their relevance to the context of the current teaching unit. Factors for assessing relevance include: the semantic relevance of the countermeasures to the dimension to be improved, the similarity of the course type or student group involved in the case, and the historical data recording the improvement effect in the case. The structured diagnostic report is self-organized, and the report generation engine automatically organizes and filters the search results according to the preset report template; The template includes structured sections such as "Problem Diagnosis", "Cause Analysis", "Improvement Suggestions", "Reference Cases" and "Recommended Resources". The engine will fill the corresponding chapters with the mapping of the dimensions to be improved to knowledge graph queries, the subgraph structure closely related to the dimensions to be improved, and the priority sorting and filtering of countermeasures and cases, and generate a personalized diagnostic report containing specific text descriptions, data references and resource links. The report output and feedback channels are established. The generated personalized diagnostic reports are pushed to the target teachers through a visual interactive platform. At the same time, the system embeds a feedback interface in the report to collect teachers' willingness to adopt the report's suggestions or the implementation effect. This feedback data can be used as input for subsequent optimization of knowledge graph relationships or case effects.

[0027] A teaching quality assessment and diagnostic system based on dynamic credibility weighting and knowledge graph includes: Multi-source data acquisition and preprocessing module: Collects and standardizes multi-source heterogeneous data from four types of data sources: teaching process, subjective evaluation, teaching background and teacher development, and quantifies text comments through natural language processing technology; Credibility Dynamic Weighted Fusion Evaluation Module: Dynamically calculates credibility weights for each data source or evaluator, adjusts weights based on cross-validation of multi-source data, and performs weighted fusion of multi-dimensional evaluation indicators to generate a credibility fusion evaluation score with confidence intervals; Teaching quality dynamic early warning module: Constructs a time-series dynamic teaching profile, monitors real-time data based on preset rules, and automatically generates and pushes early warning events when early warning conditions are triggered; Personalized diagnostic report generation module: Based on the credible fusion assessment score, it identifies teaching strengths and areas for improvement, and automatically generates a structured diagnostic report by combining the teaching improvement knowledge graph; Visual interactive platform: Displays dynamic profiles, early warning information, assessment results and diagnostic reports, and provides a human-computer interaction interface.

[0028] In one embodiment, the credibility dynamic weighted fusion evaluation module includes a basic credibility calculation unit, a dynamic weight adjustment unit, and a fusion evaluation score calculation unit. The basic credibility calculation unit calculates the basic credibility weights of various data sources based on the evaluator's historical consistency, the student's evaluation discrimination and seriousness, and the completeness of teaching process data. The dynamic weight adjustment unit adjusts the weight of each data source in the current teaching unit in real time based on the consistency, correlation or conflict between data through a multi-source data cross-validation mechanism. The fusion evaluation score calculation unit performs weighted fusion of standardized multi-dimensional evaluation indicators based on dynamic weights, and outputs a credible fusion evaluation score with a credibility level.

[0029] In one embodiment, the dynamic early warning module for teaching quality includes a profile building unit, a rule engine unit, and an early warning generation and push unit: The portrait construction unit connects the multi-dimensional credible integrated evaluation scores of each teaching unit in chronological order to form a visually representable dynamic teaching status trend chart. The rule engine unit embeds four types of early warning rules: threshold, trend, comparison, and conflict, which automatically match and judge the real-time incoming teaching data stream. The warning generation and push unit automatically generates warning events when a rule is triggered, and pushes the warning information to teaching administrators and teachers' terminals in real time through a message queue or interface.

[0030] In one embodiment, the personalized diagnostic report generation module includes a problem identification unit, a knowledge graph query unit, and a report synthesis unit: The problem identification unit analyzes the scores and trends of the credible fusion evaluation score in each dimension, and identifies the dimensions with significant advantages and dimensions that need improvement. The knowledge graph query unit is linked to built-in or external teaching improvement knowledge graphs, and retrieves related cause analyses, improvement strategies, success cases and teaching resources based on the identified problems. The report synthesis unit automatically organizes the search results into a structured text report that includes problem diagnosis, specific suggestions, reference cases and resource links, and supports output in document or visualization form.

[0031] All data obtained in this invention has been authorized by the user.

[0032] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A teaching quality assessment and diagnostic method based on dynamic credibility weighting and knowledge graphs, characterized in that, Includes the following steps: Multi-source heterogeneous data collection and standardized preprocessing: Collect teaching process data, subjective evaluation data, teaching background data and teacher development data, clean, align and standardize the collected data, quantify the text comments, extract feature vectors, and form a standardized dataset; Credibility-based dynamic weighted fusion assessment: Based on the standardized dataset, the basic credibility of each data source or evaluator is calculated. By analyzing the consistency, correlation, or conflict between the assessment indicators reflected by different data sources within the same teaching unit, the cross-validation results of the multi-source data of the current teaching unit are obtained. The basic credibility weights are dynamically adjusted according to the cross-validation results. The multi-dimensional assessment indicators are weighted and fused according to the adjusted dynamic weights to generate a credibility fusion assessment score with confidence intervals or credibility levels. Construction and real-time early warning of dynamic teaching quality profile: The trusted fusion evaluation scores are connected in time sequence to construct a dynamic teaching profile. Based on preset business rules, the real-time incoming data is monitored. When threshold early warning, trend early warning, comparison early warning or conflict early warning is triggered, an early warning event is generated and pushed to the specified object. Intelligent generation of personalized diagnostic reports: Based on predefined identification rules, the credible fusion evaluation score is analyzed to obtain the significant advantages and dimensions to be improved of the teaching unit. The teaching improvement knowledge graph is linked, and matching improvement cases, teaching strategies and related resources are retrieved according to the identified dimensions to generate a structured personalized diagnostic report.

2. The method according to claim 1, characterized in that, The teaching process data includes student video viewing time, assignment submission time, forum posts, chapter test scores, and login frequency behavior logs obtained from the online learning platform. The subjective evaluation data includes quantitative scores and text comments from students, supervisors, and peers obtained from the teaching evaluation system, as well as class observation records obtained from the class observation system. The teaching background data includes course attributes, student grades, class sizes, and academic performance data obtained from the academic affairs system. The teacher development data includes teaching training records, teaching reform projects, and awards obtained from teacher files.

3. The method according to claim 1, characterized in that, The calculation of the basic credibility of each data source or evaluator includes: For supervisory and peer evaluations, the basic credibility is calculated based on the correlation coefficient of long-term consistency between their historical evaluations and the comprehensive benchmark. For student evaluations of teaching, the basic credibility is calculated by comprehensively considering the consistency of all student evaluations, the length of the evaluation text, and the degree of dispersion with other student evaluations. For teaching process data, the basic reliability is calculated based on the completeness and stability of the data.

4. The method according to claim 1, characterized in that, The dynamic adjustment of the basic confidence weights based on the cross-validation results specifically includes: When the performance of multiple data sources within the same teaching unit corroborates each other, the temporary weight of the corresponding data source is increased. When the evaluation results of the data source conflict significantly with the conclusions of other data sources, a verification mechanism is triggered and the dynamic weight of the data source is reduced. The cross-validation covers the correlation analysis between teaching process data, subjective evaluation data, teaching background data, and teacher development data.

5. The method according to claim 1, characterized in that, The preset business rules include threshold warning rules, trend warning rules, comparison warning rules, and conflict warning rules: The threshold warning rule means that a warning will be issued if a student's online learning time, timely homework submission rate, and frequency of classroom interaction exceed a preset fluctuation range for several consecutive weeks. The trend warning rule is defined as identifying significant changing trends in dimensional indicators, including significant downward trends in positive indicators and significant upward trends in negative indicators. The comparison warning rule means that a warning is triggered when the score of a certain dimension is consistently lower than the preset range of the average level of similar courses in the same college. The conflict warning rule means that when there are high contradictions in multi-source data within the same teaching unit, an abnormality in the evaluation process will be indicated and a warning will be generated.

6. The method according to claim 1, characterized in that, The teaching improvement knowledge graph is a structured knowledge base that stores the relationships between teaching problems, causes, countermeasures, success cases, and related resources. The report generation engine, based on the problem dimensions identified by the trusted fusion evaluation score, retrieves matching countermeasures and cases from the knowledge graph and automatically organizes and generates a personalized diagnostic report containing specific improvement suggestions, reference cases, and resource links.

7. A teaching quality assessment and diagnostic system based on dynamic credibility weighting and knowledge graph, characterized in that, include: Multi-source data acquisition and preprocessing module: Collects and standardizes multi-source heterogeneous data from four types of data sources: teaching process, subjective evaluation, teaching background and teacher development, and quantifies text comments through natural language processing technology; Credibility Dynamic Weighted Fusion Evaluation Module: Dynamically calculates credibility weights for each data source or evaluator, adjusts weights based on cross-validation of multi-source data, and performs weighted fusion of multi-dimensional evaluation indicators to generate a credibility fusion evaluation score with confidence intervals; Teaching quality dynamic early warning module: Constructs a time-series dynamic teaching profile, monitors real-time data based on preset rules, and automatically generates and pushes early warning events when early warning conditions are triggered; Personalized diagnostic report generation module: Based on the credible fusion assessment score, it identifies teaching strengths and areas for improvement, and automatically generates a structured diagnostic report by combining the teaching improvement knowledge graph; Visual interactive platform: Displays dynamic profiles, early warning information, assessment results and diagnostic reports, and provides a human-computer interaction interface.

8. The system according to claim 7, characterized in that, The credibility dynamic weighted fusion evaluation module includes a basic credibility calculation unit, a dynamic weight adjustment unit, and a fusion evaluation score calculation unit. The basic credibility calculation unit calculates the basic credibility weights of various data sources based on the evaluator's historical consistency, the student's evaluation discrimination and seriousness, and the completeness of teaching process data. The dynamic weight adjustment unit adjusts the weight of each data source in the current teaching unit in real time based on the consistency, correlation or conflict between data through a multi-source data cross-validation mechanism. The fusion evaluation score calculation unit performs weighted fusion of standardized multi-dimensional evaluation indicators based on dynamic weights, and outputs a credible fusion evaluation score with a credibility level.

9. The system according to claim 7, characterized in that, The dynamic early warning module for teaching quality includes a profile building unit, a rule engine unit, and an early warning generation and push unit: The portrait construction unit connects the multi-dimensional credible integrated evaluation scores of each teaching unit in chronological order to form a visually representable dynamic teaching status trend chart. The rule engine unit embeds four types of early warning rules: threshold, trend, comparison, and conflict, which automatically match and judge the real-time incoming teaching data stream. The warning generation and push unit automatically generates warning events when a rule is triggered, and pushes the warning information to teaching administrators and teachers' terminals in real time through a message queue or interface.

10. The system according to claim 7, characterized in that, The personalized diagnostic report generation module includes a problem identification unit, a knowledge graph query unit, and a report synthesis unit: The problem identification unit analyzes the scores and trends of the credible fusion evaluation score in each dimension, and identifies the dimensions with significant advantages and dimensions that need improvement. The knowledge graph query unit is linked to a built-in or external teaching improvement knowledge graph, and retrieves related cause analysis, improvement strategies, success cases and teaching resources based on the identified problems; The report synthesis unit automatically organizes the search results into a structured text report that includes problem diagnosis, specific suggestions, reference cases, and resource links.

Images