Multimodal Data-Driven Education Evaluation Method and System

Abstract

The present invention relates to the technical field of data analysis, and particularly to a multi-modal data-driven education evaluation method and system. The method includes the following steps: collecting multi-modal education data; performing global integration and incremental iterative mapping processing on the education data based on the multi-modal education data to generate a global mapping incremental education matrix; performing structured and unstructured education feature analysis according to the global mapping incremental education matrix to generate education feature data; obtaining historical education evaluation data; performing education hierarchy evaluation index analysis according to the education feature data and the historical education evaluation data to generate education hierarchy evaluation index data; designing an intelligent education evaluation hierarchy relationship model based on the education hierarchy evaluation index data; and performing intelligent education evaluation processing on the education feature data based on the intelligent education evaluation hierarchy relationship model to generate intelligent education evaluation data. The present invention realizes efficient and accurate intelligent education evaluation through the drive analysis of education multi-modal data.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular, to a multi-modal data-driven education evaluation method and system. Background Art

[0002] With the rapid development of information technology, the education industry has gradually entered the era of intelligence and data-driven. Educational evaluation is an important means to optimize teaching methods and improve students' learning effects. Through educational evaluation, the learning performance, knowledge mastery, etc. of students are objectively analyzed, providing data support for teachers to improve teaching strategies, formulating scientific education policies, optimizing the allocation of educational resources, and improving the overall educational quality. The methods of educational evaluation mostly rely on examination scores, teacher evaluations, or questionnaires, and it is difficult to comprehensively reflect the learning process and ability development of students. With the development of technologies such as artificial intelligence and big data analysis, more accurate and personalized evaluation results are provided through intelligent analysis of education-related behaviors, thus promoting the intelligent and scientific development of education. However, the existing multi-modal data-driven education evaluation methods are difficult to effectively process multi-modal education data. The diversity of education data makes feature extraction and modeling relatively complex. How to effectively analyze structured and unstructured features and construct a reasonable evaluation index system is the key factor affecting the effect of education evaluation, and usually based on a fixed index system, it is difficult to adapt to the changes in education models and personalized needs. Summary of the Invention

[0003] Based on this, the present invention provides a multi-modal data-driven education evaluation method and system to solve at least one of the above technical problems.

[0004] To achieve the above object, a multi-modal data-driven education evaluation method includes the following steps:

[0005] Step S1: Use a preset educational administration system API interface to collect multi-modal education data, and generate multi-modal education data; perform global integration and incremental iterative mapping processing on the education data based on the multi-modal education data to generate a global mapping incremental education matrix;

[0006] Step S2: Perform structured and unstructured education feature analysis according to the global mapping incremental education matrix to generate education feature data;

[0007] Step S3: Obtain historical education evaluation data; perform education level evaluation index analysis according to the education feature data and the historical education evaluation data to generate education level evaluation index data; design a tree-level mapping relationship between educational activities and education evaluation based on the education level evaluation index data to obtain an intelligent education evaluation level relationship model;

[0008] Step S4: Perform educational intelligent evaluation processing on educational feature data based on the intelligent teaching evaluation hierarchical relationship model, generate educational intelligent evaluation data, and transmit the educational intelligent evaluation data to the terminal to execute the intelligent feedback operation of teaching evaluation.

[0009] Further, step S1 includes the following steps:

[0010] Step S11: Use the preset API interface of the educational administration system to collect multi-modal educational data and generate multi-modal educational data;

[0011] Step S12: Design an educational logical structure relationship matrix based on the multi-modal educational data and generate an educational logical structure relationship matrix;

[0012] Step S13: Design an educational logical structure iteration strategy based on the educational logical structure relationship matrix and generate an educational logical structure iteration strategy;

[0013] Step S14: Collect global educational data according to the multi-modal educational data, generate global educational data, and transmit the global educational data to the educational logical structure relationship matrix for preliminary mapping processing of educational data to generate a global mapping educational matrix;

[0014] Step S15: Perform instant update of incremental educational data according to the multi-modal educational data, generate instant incremental educational data, and transmit the instant incremental educational data to the global mapping educational matrix through the educational logical structure iteration strategy for iterative mapping processing of incremental educational data to generate a global mapping incremental educational matrix.

[0015] Further, step S13 includes the following steps:

[0016] Step S131: Analyze the educational logical structure features based on the educational logical structure relationship matrix and generate educational logical structure feature data;

[0017] Step S132: Analyze the iterative requirements of the educational logical structure based on the educational logical structure feature data and generate educational logical structure iterative requirement data;

[0018] Step S133: Analyze the standardization requirements of educational data according to the educational logical structure feature data and the data types of multi-modal educational data, and generate educational data standardization requirement data;

[0019] Step S134: Design an educational logical structure iteration strategy through the educational logical structure iterative requirement data and the educational data standardization requirement data, and generate an educational logical structure iteration strategy.

[0020] Further, step S2 includes the following steps:

[0021] Step S21: Conduct an educational structured feature analysis on the global mapping incremental education matrix to generate educational structured feature data;

[0022] Step S22: Conduct an analysis of educational unstructured key behavior events based on the global mapping incremental education matrix to obtain educational unstructured key behavior event data;

[0023] Step S23: Design an educational unstructured feature mapping rule based on the educational unstructured key behavior event data, and use the educational unstructured feature mapping rule to perform an educational unstructured feature mapping process on the global mapping incremental education matrix to generate educational unstructured feature data;

[0024] Step S24: Conduct an educational feature integration process on the educational structured feature data and the educational unstructured feature data to generate educational feature data.

[0025] Further, step S22 includes the following steps:

[0026] Step S221: Conduct an educational activity behavior analysis based on the global mapping incremental education matrix to generate educational activity behavior data; and perform an educational activity behavior object detection through the educational activity behavior data to generate educational activity behavior object data;

[0027] Step S222: Use the educational activity behavior data to perform an educational unstructured behavior analysis on the global mapping incremental education matrix to generate educational unstructured behavior analysis data;

[0028] Step S223: Conduct an educational unstructured behavior pattern analysis based on the educational activity behavior object data and the educational unstructured behavior analysis data to generate educational unstructured behavior pattern data;

[0029] Step S224: Conduct an educational unstructured key behavior event analysis on the educational unstructured behavior analysis data according to the educational unstructured behavior pattern data to obtain educational unstructured key behavior event data.

[0030] Further, step S224 includes the following steps:

[0031] Design an educational key behavior event detection threshold through the educational unstructured behavior pattern data;

[0032] Conduct a behavior pattern time series analysis based on the educational unstructured behavior pattern data to generate educational unstructured behavior pattern time series data, and design an educational key behavior event time series detection threshold through the educational unstructured behavior pattern time series data;

[0033] Extract educational unstructured key behavior event data by extracting educational unstructured behavior analysis data that meets the educational key behavior event detection threshold and the corresponding educational key behavior event time series detection threshold.

[0034] Further, step S3 includes the following steps:

[0035] Step S31: Perform educational feature dimension clustering processing on educational feature data to generate educational feature dimension clustering data;

[0036] Step S32: Perform educational evaluation index classification processing based on the educational feature dimension clustering data to generate educational evaluation classification index data;

[0037] Step S33: Design educational evaluation category indicators and corresponding educational evaluation category subset indicators based on the educational evaluation classification index data;

[0038] Step S34: Obtain historical educational evaluation data;

[0039] Step S35: Perform educational index multi-level quantization evaluation characteristic analysis on the educational evaluation category subset indicators according to the historical educational evaluation data to generate educational index multi-level quantization evaluation characteristic data;

[0040] Step S36: Perform educational level evaluation index integration processing through the corresponding educational evaluation category indicators, educational evaluation category subset indicators, and educational index multi-level quantization evaluation characteristic data to generate educational level evaluation index data;

[0041] Step S37: Design a tree-level mapping relationship between educational activities and educational evaluations based on the educational level evaluation index data to obtain an intelligent educational evaluation level relationship model.

[0042] Further, step S33 includes the following steps:

[0043] Step S331: Design educational evaluation category indicators through the educational evaluation classification index data;

[0044] Step S332: Perform factor influence factor analysis on the classification indicators according to the educational evaluation classification index data to generate classification indicator factor influence factor data;

[0045] Step S333: Design educational evaluation category subset indicators through the classification indicator factor influence factor data corresponding to the educational evaluation category indicators.

[0046] Further, step S4 includes the following steps:

[0047] Step S41: Analyze the educational evaluation requirement tasks for the educational feature data based on the global educational data and the instant incremental educational data, and generate educational evaluation requirement task data;

[0048] Step S42: Dynamically adjust and optimize the model of the educational evaluation requirement tasks for the intelligent educational evaluation hierarchical relationship model according to the educational evaluation requirement task data, and generate an optimized intelligent educational evaluation hierarchical relationship model;

[0049] Step S43: Transmit the educational feature data to the optimized intelligent educational evaluation hierarchical relationship model for intelligent evaluation processing of the educational hierarchy, and generate intelligent evaluation data for the educational hierarchy;

[0050] Step S44: Conduct collaborative analysis of the intelligent educational evaluation based on the intelligent evaluation data for the educational hierarchy, generate intelligent educational evaluation data, and transmit the intelligent educational evaluation data to the terminal to execute the intelligent feedback operation for educational evaluation.

[0051] This specification provides a multi-modal data-driven educational evaluation system for executing the multi-modal data-driven educational evaluation method as described above. The multi-modal data-driven educational evaluation system includes:

[0052] An educational data collection module, which is used to collect multi-modal educational data by using a preset educational administration system API interface, and generate multi-modal educational data; conduct global integration and incremental iterative mapping processing of the educational data based on the multi-modal educational data to generate a global mapping incremental educational matrix;

[0053] An educational feature analysis module, which is used to conduct structured and unstructured educational feature analysis based on the global mapping incremental educational matrix, and generate educational feature data;

[0054] An educational evaluation hierarchical relationship model design module, which is used to obtain historical educational evaluation data; conduct educational hierarchy evaluation index analysis based on the educational feature data and the historical educational evaluation data, and generate educational hierarchy evaluation index data; design a tree-shaped hierarchical mapping relationship between educational activities and educational evaluations based on the educational hierarchy evaluation index data to obtain an intelligent educational evaluation hierarchical relationship model;

[0055] An intelligent educational evaluation feedback module, which is used to conduct intelligent educational evaluation processing on the educational feature data based on the intelligent educational evaluation hierarchical relationship model, generate intelligent educational evaluation data, and transmit the intelligent educational evaluation data to the terminal to execute the intelligent feedback operation for educational evaluation.

[0056] The beneficial effects of this application are as follows. The present invention collects multi-modal education data through a preset educational administration system API interface, constructs a complete education data collection and processing mechanism, and realizes efficient and dynamic data management, which not only depends on static data (such as exam scores, attendance records, etc.), but also includes classroom interaction, video analysis, text records, etc., achieving more comprehensive education data collection, thus providing a more accurate basis for education evaluation. And through the educational logic structure relationship matrix, it is used to construct the logical relationship framework of education data, making the data organization more reasonable and better able to reflect the causal relationship and relevance in the education process. Through the design of the educational logic structure iteration strategy, the system can continuously optimize its own data model according to the real-time collected data, realizing adaptive data update. This incremental iteration method based on the logical structure makes the data mapping process more accurate and can effectively avoid the inefficiency problems caused by data redundancy and strong heterogeneity in management. By constructing the global mapping incremental education matrix, the integrity and real-time nature of the data are ensured. By adopting the combination of global education data collection and instant incremental update, new data can be quickly incorporated into the system and intelligently mapped and associated based on the existing educational logic relationship, greatly improving the real-time nature of the data, enabling the education evaluation system to quickly adapt to the changes in the education environment, and providing high-quality data support for subsequent education feature analysis and evaluation. Through educational structured feature analysis, structured data such as students' grades, homework completion situations, and classroom interaction situations can be accurately extracted, enabling education evaluation to be analyzed based on objective and quantifiable data. Secondly, in the analysis of educational unstructured key behavior events, valuable information is extracted from unstructured data such as classroom videos, voice data, and text records. The educational unstructured feature mapping rules are introduced to convert unstructured data into structured information available for analysis. For example, through the behavior pattern detection threshold and the time series detection threshold, the abnormal behaviors of students in the classroom are automatically identified and mapped into key behavior event data, which not only improves the utilization rate of unstructured data but also enhances the ability to identify hidden problems in the education process. Through the integration and processing of educational features, structured data and unstructured data are deeply fused, enabling the education evaluation system to consider both learning outcomes and learning processes simultaneously, making the education evaluation more accurate and comprehensive, and being able to dynamically adjust according to real-time data, avoiding the problem of one-sided education evaluation. Through the clustering processing of educational feature dimensions, educational feature data is automatically identified and classified, making the relationship between different educational features clearer. Using historical education evaluation data for multi-level quantitative analysis makes the education evaluation system more accurate. Through historical data analysis, different education evaluation indicators are quantified, thereby establishing a more scientific evaluation system and making the evaluation more detailed and fair.Design an evaluation system for multi-level indicators through a tree-like hierarchical mapping relationship, expanding from a single dimension to a multi-dimensional level, forming an intelligent evaluation system covering multiple dimensions such as curriculum teaching, learning process, and learning outcomes, which can effectively reflect the complexity and dynamics of educational activities. Based on the intelligent teaching evaluation hierarchical relationship model, intelligent evaluation processing of educational feature data is realized, and through the intelligent feedback mechanism of teaching evaluation, instant feedback and optimization adjustment of educational evaluation results are achieved. Through the analysis of educational evaluation requirement tasks, the educational evaluation criteria can be dynamically adjusted according to different educational goals and application scenarios, and personalized educational evaluation requirement tasks can be automatically generated based on global educational data and incremental educational data to ensure the accuracy and applicability of the evaluation system. By adopting the dynamic optimization mechanism of the intelligent teaching evaluation hierarchical relationship model, it can be adjusted according to real-time educational data to ensure that the evaluation model can continuously adapt to the changing educational environment. For example, according to the newly collected learning behavior data, the weight distribution of the hierarchical relationship model is optimized to ensure the reasonable weights of different evaluation indicators, thereby improving the accuracy and fairness of evaluation results. In addition, through the collaborative analysis of educational intelligent evaluation of different educational entities, the evaluation mechanism is optimized to achieve the interpretability and operability of educational evaluation results, transmit the educational intelligent evaluation data to terminal devices, realize the seamless connection of intelligent feedback and teaching optimization, and achieve intelligent education management driven by data.

[0057] Therefore, the multi-modal data-driven educational evaluation method of the present invention constructs a multi-modal educational data collection mechanism based on a preset API interface of the educational administration system, which can integrate structured and unstructured data, including course arrangements, exam scores, classroom interactions, voice texts, video images, etc., to achieve the efficient fusion of multi-source heterogeneous educational data. By constructing a global mapping incremental educational matrix, educational data is dynamically updated to ensure the integrity and timeliness of educational evaluation data, overcoming the problem of difficult data integration in the prior art. By using structured and unstructured feature extraction methods, based on the global mapping incremental educational matrix, the educational behaviors are analyzed structurally, and through the unstructured key behavior event detection mechanism, the unstructured feature information such as classroom teaching and students' learning status is mined, and the educational behavior data that is difficult to cover by educational evaluation is obtained, and the key factors affecting learning quality can also be effectively identified, thereby enhancing the comprehensiveness and scientificity of educational evaluation. The introduction of the intelligent teaching evaluation hierarchical relationship model divides educational evaluation into different levels, and the model is optimized based on historical educational evaluation data to adapt to the needs of different educational environments, supports the dynamic adjustment of educational evaluation indicators, realizes personalized and multi-level educational evaluation, and solves the problem of lack of dynamic adaptability in the evaluation system in the prior art. Based on the educational intelligent evaluation data, personalized evaluation results are provided, and real-time adjustment of teaching strategies is supported, realizing dynamic evaluation and instant feedback in the educational process, enhancing the pertinence and interactivity of teaching, and further improving the efficiency of educational management and teaching optimization. Brief Description of the Drawings

[0058] Figure 1 This is a schematic diagram of the step process of an education evaluation method driven by multimodal data according to the present invention;

[0059] Figure 2 is Figure 1 a detailed implementation step process diagram of step S2 in

[0060] Figure 3 is Figure 1 a detailed implementation step process diagram of step S3 in

[0061] The realization, functional characteristics, and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the accompanying drawings. Specific Embodiments

[0062] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0063] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0064] It should be understood that although the terms "first", "second", etc. may be used here to describe each unit, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed associated items.

[0065] To achieve the above object, please refer to Figures 1 to 3 , the present invention provides an education evaluation method driven by multimodal data. In the embodiments of the present invention, please refer to Figure 1 shown, which is a schematic diagram of the step process of an education evaluation method driven by multimodal data according to the present invention. The education evaluation method driven by multimodal data includes the following steps:

[0066] Step S1: Use a preset educational administration system API interface to collect multimodal educational data and generate multimodal educational data; perform global integration and incremental iterative mapping processing on the educational data based on the multimodal educational data to generate a global mapping incremental educational matrix;

[0067] In the embodiments of the present invention, multi-modal education data is collected by using a preset educational administration system API interface. By calling the existing API interface in the educational administration system, data in multiple dimensions such as teaching activities, student behavior, academic performance, and teacher feedback are collected. The specific API interface settings may include the access method of the API (such as RESTful API), data type (such as JSON format), data fields (such as student ID, course ID, grades, assignment numbers, etc.). An educational logic structure relationship matrix is designed based on the multi-modal education data, the multi-modal education data is structured, and various dimensional features related to educational activities are extracted, such as student learning progress, classroom participation, assignment completion, etc. Then, through mathematical modeling (such as graph theory or matrix operations), the relationships between the dimensional features are mapped into matrix form to generate an educational logic structure relationship matrix. For example, the educational logic structure relationship matrix can be composed of the interaction relationships between students and courses, where the rows of the matrix represent students, the columns represent courses, and the elements of the matrix represent the performance of students in each course (such as grades, participation), and the matrix generation process uses matrix algorithms (such as SVD decomposition) for relationship mapping to optimize the correlation of each dimensional feature. An educational logic structure iterative strategy is designed based on the educational logic structure relationship matrix to generate an educational logic structure iterative strategy. According to the characteristics of the educational logic structure relationship matrix, an iterative strategy is designed so that educational data can be dynamically updated at different stages, using adaptive learning algorithms (such as reinforcement learning or iterative optimization algorithms). Adjust the educational structure relationship according to the historical data feedback. The design of the iterative strategy requires automatically adjusting the weights of various indicators according to the dynamic changes of multi-modal data to achieve intelligent iterative update of educational data at different stages. Global educational data is collected according to the multi-modal education data. First, global educational data is collected from each educational subsystem (such as student information management, course management, grade management, etc.), and these data are aggregated to a unified data platform through an API interface. And the global educational data is transmitted to the educational logic structure relationship matrix for preliminary mapping processing of educational data. After data collection, the global educational data is mapped to the educational logic structure relationship matrix according to the set logical structure. For example, a weighted sum model based on weights can be used, combined with information such as the specific course grades and assignment completion degrees of each student, to generate a global mapping education matrix. According to the changes in multi-modal education data (such as student grades, learning activity data, classroom feedback, etc.), newly added or updated data is obtained in real time and converted into incremental educational data. At this time, the acquisition of incremental data can be carried out through a real-time monitoring system, using timestamps to mark the new data and comparing it with the previous historical data to ensure the timeliness and accuracy of the data. These incremental data are processed through the educational logic structure iterative strategy. This strategy includes steps such as data integration, deduplication, and standardization to ensure the consistency and integrity of the incremental data.Use iterative algorithms (such as incremental training models or incremental learning algorithms) to combine incremental educational data with the existing global mapping educational matrix, and gradually update the various educational indicators in the global mapping matrix. Through weighted average method or incremental matrix decomposition technology for data mapping processing, new data can be quickly and accurately incorporated into the existing model to generate a global mapping incremental educational matrix.

[0068] Step S2: Conduct structured and unstructured educational feature analysis based on the global mapping incremental educational matrix to generate educational feature data;

[0069] In the embodiments of the present invention, for the educational structured feature analysis of the global mapping incremental education matrix, first, various structured data are extracted from the global mapping incremental education matrix, which usually includes students' basic information (such as grade, class, course selection, etc.), exam scores, attendance rate, homework completion situation, etc. These data constitute the structured features in the educational process. To conduct feature analysis, the principal component analysis (PCA) algorithm is used to reduce the dimension of the extracted educational data, so as to facilitate the identification of the main influencing factors and their relative weights. Through this step, it can be identified which educational features (such as course grades, homework completion degree, etc.) are most relevant to key indicators such as students' learning effects and participation. For the analysis of educational unstructured key behavior events based on the global mapping incremental education matrix, first, unstructured data are identified from the incremental educational data, such as students' classroom interaction behaviors, learning discussions on social media, learning duration in the online learning platform, etc. These unstructured data need to be extracted and classified through natural language processing (NLP) techniques or behavior analysis models. An event detection model (such as an LSTM model based on deep learning) is used to analyze the key behavior events of students during the learning process, such as long-term unfinished homework or high-frequency interactive learning group discussions. By setting the detection threshold of the behavior event, if a student fails to complete homework twice consecutively in a learning module, the system regards this behavior as a low-performance behavior and marks this behavior. The educational unstructured key behavior event data will provide more targeted intervention signals for subsequent educational decisions. For designing the educational unstructured feature mapping rules based on the educational unstructured key behavior event data, first, the identified unstructured key behavior events (such as low-performance students, low course participation, low academic interaction frequency, etc.) need to be classified and labeled. Classification algorithms (such as support vector machine SVM or decision tree algorithm) are used to further analyze the features of educational unstructured behavior events, and key factors affecting students' performance are extracted, such as homework submission frequency, online discussion participation, interaction frequency with teachers, etc. Feature mapping rules are designed to map these unstructured behavior features into the global mapping incremental education matrix. For example, when the homework submission frequency of a student within a certain stage is lower than the set threshold, the behavior feature of this student can be marked as "low participation" through the mapping rule. In addition, when designing the rules, the consistency of educational background data is considered to ensure that the mapping rules can adapt to data changes in different educational scenarios. The setting of the mapping rules includes: frequency threshold, behavior frequency analysis window, and mapping target feature dimension. The mapping rules are applied to the global mapping incremental education matrix to generate educational unstructured feature data. For the integration processing of educational structured feature data and educational unstructured feature data, the extracted educational structured feature data (such as students' exam scores, attendance, homework completion rate, etc.) are integrated with the educational unstructured feature data (such as students' classroom interaction frequency, online learning behavior, social media discussion participation, etc.).Use machine learning models such as the weighted average method, principal component analysis (PCA), or multi-layer perceptron (MLP) for feature integration, merge the two types of feature data according to certain weights, so as to obtain comprehensive educational feature data. The integrated educational feature data will cover multiple dimensions such as students' academic performance, behavior patterns, learning attitudes, etc., forming a complete educational feature dataset. The results of this step can provide support for subsequent educational decisions. For example, it can help teachers design personalized teaching strategies or optimize learning path recommendations, etc.

[0070] Step S3: Obtain historical educational evaluation data; conduct an analysis of educational level evaluation indicators based on the educational feature data and historical educational evaluation data to generate educational level evaluation indicator data; design a tree-level mapping relationship between educational activities and educational evaluations based on the educational level evaluation indicator data to obtain an intelligent teaching evaluation level relationship model;

[0071] In the embodiments of the present invention, clustering analysis is performed on educational feature data with the aim of discovering potential patterns in the data and grouping related features, thereby simplifying the subsequent evaluation and analysis process. The K-means clustering algorithm is used for clustering the educational feature dimensions. Through this clustering process, educational feature dimension clustering data is generated, where each clustering result corresponds to a group of students with similar educational features. Classification processing of educational evaluation indicators is performed based on the educational feature dimension clustering data. According to the clustering results of different feature dimensions, the learning behaviors and performances of each type of student are analyzed and mapped into a preset educational evaluation index system. For example, for the high-achievement group, indicators such as "academic performance" and "exam pass rate" are mainly considered; while for the low-participation group, more attention is paid to behavioral indicators such as "learning duration" and "interaction frequency". The classification of educational evaluation indicators can be carried out through methods such as hierarchical clustering or decision trees. Using a decision tree model, classification indicators are generated based on the clustering results of students and various feature data. The set classification basis includes key dimensions such as students' participation, academic performance, and learning behaviors. Through this process, educational evaluation classification index data is generated, providing structured data support for the subsequent evaluation model. Based on the educational evaluation classification index data, educational evaluation category indicators and corresponding subset indicators are further designed. According to the characteristics of each group, suitable evaluation categories are designed. For example, for the academic performance group, the "academic ability" category is designed, and the subset indicators below can include "exam scores", "assignment completion rate", "academic contribution degree", etc.; for the behavioral performance group, the "learning attitude" category is designed, and the subset indicators can include "classroom participation", "timely assignment submission rate", "after-class feedback", etc. The weights of each category indicator and subset indicator are determined through expert review combined with statistical analysis methods. For example, the expert scoring method combined with the AHP (Analytic Hierarchy Process) model is used for weight allocation of evaluation categories to ensure the rationality and comparability of each category indicator, so as to generate educational evaluation category indicators and subset indicators, providing a basis for educational level evaluation and subsequent intelligent evaluation. Historical educational evaluation data is obtained from educational data sources or systems. Historical educational evaluation data can include students' grade records, classroom performance data, assignment feedback, educational evaluation records, etc. Relevant data is extracted from the database through the API interface of the educational administration system. A multi-level scoring model (such as the Analytic Hierarchy Process AHP) is used to perform multi-level quantitative evaluation on each subset indicator. By using a multi-level scoring system, the evaluation of each dimension is divided into several levels. For example, the historical educational data is standardized so that different categories of data can be compared uniformly. According to the nature of each subset indicator, a quantitative standard is set. The Analytic Hierarchy Process (AHP) combined with expert review is used to weight the relative importance of each subset indicator to obtain the weight of each indicator in the overall score. In this way, a quantitative evaluation model for each subset indicator is constructed.Combined with historical education evaluation data and quantitative criteria, perform multi-level quantitative characteristic analysis on the subset indicators of each education evaluation category to generate multi-level quantitative evaluation characteristic data. Through the comprehensive analysis of education evaluation category indicators, education evaluation category subset indicators, and multi-level quantitative evaluation characteristic data of education indicators, conduct the integration process of education level evaluation indicators, and finally generate education level evaluation indicator data. Adopt technical means such as weighted average method, hierarchical weighted method, or factor analysis method to integrate the evaluation data of each category indicator and subset indicator to generate a comprehensive evaluation indicator. For the comprehensive evaluation of students, multiple subset indicators (such as exam scores, homework completion, and class participation) under the "academic performance" category can be weighted according to preset weights to generate the comprehensive evaluation data for the "academic performance" category. Collect the quantitative data of education evaluation category indicators, education evaluation category subset indicators, and education indicators. These data can be sourced from historical education evaluation data or obtained through a real-time data acquisition system. According to the relative importance of education evaluation categories and subset indicators, determine the weight of each indicator through expert review or statistical methods (such as Analytic Hierarchy Process AHP). Weight the quantitative data of each subset indicator according to the weight, calculate the comprehensive evaluation data for each education category (such as academic performance, social practice, etc.), and finally form the overall education level evaluation indicator data. Based on the education level evaluation indicator data, design the tree-level mapping relationship between education activities and education evaluation to form a hierarchical relationship model of an intelligent education evaluation system (intelligent education evaluation). This model is used to automatically analyze all aspects of education activities and optimize and adjust according to the evaluation data at each level. Design a tree structure, decompose different levels in the education evaluation process into different nodes, and map the evaluation data of each node according to the education level evaluation indicator data. Automatically construct the hierarchical mapping relationship model between education activities and evaluation through intelligent algorithms (such as Analytic Hierarchy Process, neural network model, etc.), automatically adjust the evaluation weight and mapping rules at each level, and optimize the implementation strategy of education activities. Verify the initially designed tree-level relationship model based on historical data, and evaluate its prediction accuracy and adaptability. If the model performs poorly, adjust the weight and mapping rules to further optimize the model.

[0072] Step S4: Based on the hierarchical relationship model of intelligent education evaluation, perform intelligent education evaluation processing on education feature data to generate intelligent education evaluation data, and transmit the intelligent education evaluation data to the terminal to execute the intelligent feedback operation of education evaluation.

[0073] In the embodiments of the present invention, through the comprehensive analysis of global education data and instant incremental education data, the demand tasks of education feature data are analyzed to identify the key demands and tasks existing in the education evaluation process. The goal of task analysis is to clarify which aspects of education features need to be concerned and optimized to ensure the effectiveness and pertinence of education evaluation. Global education data and instant incremental education data are collected, and the data sources include student achievement data, learning behavior data, teacher feedback data, student participation data, etc. Global education data usually includes historical education data, while instant incremental education data includes recently generated data, which helps to timely reflect the changes in the education process, extract the features required for education evaluation, and identify the key demands in the education process. According to the analysis results of education feature data, education evaluation demand task data is generated. The demand task data includes task objectives, task priorities, task execution times, and other contents. According to the education evaluation demand task data, the intelligent education evaluation hierarchical relationship model is dynamically adjusted and optimized. The intelligent education evaluation hierarchical relationship model analyzes and evaluates according to different dimensions of education data through a multi-level evaluation mechanism. Through dynamic adjustment and optimization, it can be ensured that the model can adapt to the demand changes in different education scenarios in real time, improving the flexibility and response ability of the model. The education evaluation demand task data is input into the intelligent education evaluation hierarchical relationship model. These data include the priorities, resource requirements, task objectives, and other contents of the tasks. The model needs to make corresponding adjustments according to these input data to meet specific education evaluation requirements. Technologies such as reinforcement learning and adaptive optimization algorithms are used to adjust and optimize the intelligent education evaluation hierarchical relationship model. The model optimization process makes dynamic adjustments according to factors such as task priorities, time limits, and data consistency. The dynamic adjustment of the intelligent education evaluation hierarchical relationship model generates an optimized model. The optimized model can respond more precisely to education evaluation requirements and has stronger adaptability. The optimized model has been dynamically adjusted according to the education evaluation demand task data to improve the adaptability and accuracy of the model in various education evaluation tasks. The education feature data that has undergone education evaluation demand task analysis is transmitted to the optimized intelligent education evaluation hierarchical relationship model. The education feature data usually includes students' learning behavior data, learning paths, learning achievements, participation in teaching activities, etc. These data reflect the performance of students in the education process and the implementation of education activities. Using the intelligent algorithms in the model, such as evaluating and defining education evaluation through fuzzy logic and learning evaluation through deep learning, multi-level intelligent evaluation processing is performed on the input education feature data. The model will evaluate education features based on evaluation criteria at different levels. According to the intelligent processing results, education hierarchical intelligent evaluation data is generated. This data contains evaluation indicators at multiple levels, such as students' comprehensive learning ability, learning interest, participation, etc., and is accompanied by corresponding recommended measures and optimization suggestions.Based on the intelligent evaluation data of educational levels, collaborative analysis algorithms (such as collaborative filtering, ensemble learning, etc.) are used to comprehensively analyze the evaluation data at different levels. The collaborative analysis method can obtain a more comprehensive and accurate evaluation result by integrating multiple evaluation dimensions. In this way, the model can discover potential relationships and patterns and provide more personalized feedback. According to the results of the collaborative analysis, educational intelligent evaluation data is generated. These data may include optimized teaching strategies, personalized learning path adjustments, teaching resource allocation suggestions, etc. The generated feedback data can help teachers and educational administrators make more accurate decisions. The generated educational intelligent evaluation data is transmitted to terminal devices, such as teacher terminals, student terminals, etc. These terminals will execute subsequent intelligent feedback operations for educational evaluation to ensure that the feedback can be timely and effectively transmitted to relevant personnel, and the educational evaluation and feedback can be automated and precise, improving the educational effect.

[0074] Further, step S1 includes the following steps:

[0075] Step S11: Use a preset educational administration system API interface to collect multi-modal educational data and generate multi-modal educational data;

[0076] Step S12: Design an educational logical structure relationship matrix based on the multi-modal educational data and generate an educational logical structure relationship matrix;

[0077] Step S13: Design an educational logical structure iterative strategy based on the educational logical structure relationship matrix and generate an educational logical structure iterative strategy;

[0078] Step S14: Collect global educational data according to the multi-modal educational data, generate global educational data, and transmit the global educational data to the educational logical structure relationship matrix for preliminary mapping processing of educational data to generate a global mapped educational matrix;

[0079] Step S15: Instantly update the incremental educational data according to the multi-modal educational data, generate instant incremental educational data, and transmit the instant incremental educational data to the global mapped educational matrix through the educational logical structure iterative strategy for iterative mapping processing of incremental educational data to generate a global mapped incremental educational matrix.

[0080] In the embodiments of the present invention, multi-modal education data is collected by using a preset educational administration system API interface. By calling the existing API interface in the educational administration system, data in multiple dimensions such as teaching activities, student behavior, academic performance, and teacher feedback are collected. For example, the system can obtain students' classroom attendance data, homework submission status, exam scores, classroom interaction records, etc. through the API interface. These data come from different educational systems such as student status management systems, exam management systems, learning platforms, etc., forming multi-modal data. The specific API interface settings can include the access method of the API (such as RESTful API), data type (such as JSON format), data fields (such as student ID, course ID, score, homework number, etc.). Design an educational logic structure relationship matrix based on the multi-modal education data, perform structured processing on the multi-modal education data, and extract various dimensional features related to educational activities, such as students' learning progress, classroom participation, homework completion, etc. Then, through mathematical modeling (such as graph theory or matrix operations), map the relationships between the dimensional features into matrix form to generate an educational logic structure relationship matrix. For example, the educational logic structure relationship matrix can be composed of the interaction relationships between students and courses, where the rows of the matrix represent students, the columns represent courses, and the elements of the matrix represent students' performance in each course (such as scores, participation). The generation process of the matrix uses matrix algorithms (such as SVD decomposition) for relationship mapping to optimize the correlation of each dimensional feature. Design an educational logic structure iterative strategy based on the educational logic structure relationship matrix to generate an educational logic structure iterative strategy. First, according to the characteristics of the educational logic structure relationship matrix, design an iterative strategy so that educational data can be dynamically updated at different stages. An adaptive learning algorithm (such as reinforcement learning or iterative optimization algorithm) is used. For example, the mathematical model for designing the iterative strategy can be: S(t + 1)=aS(t)+(1 - a)R(t), where S(t) is the educational structure relationship matrix at the current stage, R(t) is the feedback educational data, a is the learning rate, the value range of the learning rate is [0, 1], t is the current teaching stage, S(t) represents the educational logic structure relationship matrix at the current stage, and S(t + 1) represents the educational logic structure relationship matrix after iteration in the next stage, representing the degree of dependence of the model on historical data. Adjust the educational structure relationship according to the historical data feedback. The design of the iterative strategy requires automatically adjusting the weights of various indicators according to the dynamic changes of the multi-modal data to achieve intelligent iterative update of educational data at different stages. Collect global educational data according to the multi-modal education data. First, collect global educational data from each educational subsystem (such as student information management, course management, score management, etc.). These data are aggregated to a unified data platform through the API interface. And transmit the global educational data to the educational logic structure relationship matrix for preliminary mapping processing of educational data. After data collection, map the global educational data to the educational logic structure relationship matrix according to the set logical structure.For example, a weighted sum model based on weights can be used to generate a global mapping education matrix by combining information such as the specific course grades and assignment completion of each student. According to the changes in multi-modal education data (such as student grades, learning activity data, classroom feedback, etc.), new or updated data can be obtained in real time and converted into incremental education data. At this time, the acquisition of incremental data can be carried out through a real-time monitoring system, using timestamps to mark the new data and comparing it with the previous historical data to ensure the real-time and accuracy of the data. For example, for an online learning platform, information such as student activity and course completion can be tracked in real time to generate incremental data. Next, these incremental data are processed through an educational logic structure iteration strategy. This strategy includes steps such as data integration, deduplication, and standardization to ensure the consistency and integrity of the incremental data. Then, an iterative algorithm (such as an incremental training model or an incremental learning algorithm) is used to combine the incremental education data with the existing global mapping education matrix and gradually update the various educational indicators in the global mapping matrix. Data mapping processing is carried out through the weighted average method or incremental matrix decomposition technology to enable the new data to be quickly and accurately incorporated into the existing model to generate a global mapping incremental education matrix. During this processing, the data update frequency is set to once per minute, the standardization error threshold of the incremental data is set to 0.01, and the weight adjustment range during matrix iteration is from 0.05 to 0.2 to ensure that the incremental data can be updated and stably incorporated into the global education matrix in a short time.

[0081] Further, step S13 includes the following steps:

[0082] Step S131: Analyze the educational logic structure features based on the educational logic structure relationship matrix to generate educational logic structure feature data;

[0083] Step S132: Analyze the iterative requirements of the educational logic structure based on the educational logic structure feature data to generate educational logic structure iterative requirement data;

[0084] Step S133: Analyze the standardization requirements of educational data based on the educational logic structure feature data and the data types of multi-modal educational data to generate educational data standardization requirement data;

[0085] Step S134: Design an educational logic structure iteration strategy through the educational logic structure iterative requirement data and the educational data standardization requirement data to generate an educational logic structure iteration strategy.

[0086] In the embodiments of the present invention, for the analysis of educational logic structure features based on the educational logic structure relationship matrix, first, feature extraction is performed on the data in the educational logic structure relationship matrix. The principal component analysis (PCA) or feature engineering method is used to analyze the key features of educational data, such as the performance correlation of students in different courses, the similarity of learning paths, etc. Then, by calculating the statistical indicators of the matrix, such as the mean, standard deviation, covariance matrix, etc., the core features representing the educational logic structure are extracted to generate educational logic structure feature data. For example, the interaction intensity between courses can be calculated based on the row and column sums of the matrix to measure the teaching influence of the courses. In addition, the network graph analysis method is adopted to model the learning relationship between students and courses as a graph structure. By calculating the centrality indicators of the nodes (such as degree centrality, betweenness centrality, etc.), the key nodes in the educational structure are identified. For example, the proportion of the explained variance of the principal component analysis is set to more than 85%, and the calculation range of the course interaction intensity is set between [0,1] to ensure the normalization of the data. For the iterative requirement analysis of the educational logic structure based on the educational logic structure feature data, dynamic change analysis is performed on the educational logic structure feature data. For example, by comparing the changes in the educational logic structure in different time periods, it is determined which feature indicators have changed significantly, and time series analysis (such as ARIMA or LSTM models) is used to predict the future educational logic change trend. Combining historical educational data, the deviation between the current educational structure features and the existing patterns is compared, and the requirement weight for iterative adjustment is calculated. For example, a change rate threshold is set. When the change rate of a certain feature indicator exceeds 10%, the requirement for logic structure adjustment is triggered. Iterative requirement data is generated according to the analysis results, including the educational feature dimensions to be optimized, adjustment strategies, and optimization directions. For example, the time series prediction window length is set to 1 day, and the sensitivity threshold for iterative requirement adjustment is set to 10%-15% to ensure the accuracy of the iterative requirement analysis and generate educational logic structure iterative requirement data. For the analysis of educational data standardization requirements based on the educational logic structure feature data and the data types of multimodal educational data, the types of educational data are classified. For example, the data is divided into numerical data (such as grades, attendance rates), categorical data (such as course types, teaching modes), text data (such as classroom feedback, homework comments), etc. Then, standardization strategies are designed for different types of data: numerical data adopts Z-score standardization (that is, subtracting the mean and then dividing by the standard deviation) to make the data mean 0 and the standard deviation 1; categorical data adopts one-hot encoding to convert categorical variables into binary vectors; text data adopts the TF-IDF method for vectorization processing to improve computability. In addition, the missing situation of each type of data is analyzed, and data is supplemented by interpolation methods or filling default values to generate educational data standardization requirement data.For example, when the missing rate is lower than 5%, mean filling is adopted; when the missing rate is higher than 20%, KNN interpolation is used to complete the data. For example, the mean setting range of Z-score standardization is [0, 1], the TF-IDF weight threshold of text data is set to 0.01, and the data missing rate threshold is set to 5% - 20%. Through the iterative requirement data of the educational logic structure and the standardized requirement data of educational data, the iterative strategy design of the educational logic structure is carried out. Combining the iterative requirement data of the educational logic structure, the educational data feature dimensions and optimization objectives to be adjusted are determined. For example, the learning recommendation mechanism is optimized for the significantly changing curriculum learning path, or the teaching intervention strategy is adjusted for low-performing students. According to the standardized requirement data of educational data, the data is converted consistently to ensure the applicability of the iterative strategy. For example, in intelligent learning path recommendation, a reinforcement learning algorithm is introduced, and the recommendation strategy is dynamically adjusted according to the standardized student learning data. The learning rate range is set between [0.001, 0.01], and the consistency check error threshold of the standardized data is set to 0.001, so that the recommendation accuracy is increased by more than 5%. In addition, an adaptive learning rate is used to optimize the iterative strategy to ensure the convergence of the model, that is, the learning rate is increased when the data changes violently to accelerate the model adjustment speed, and the learning rate is decreased when the data changes smoothly to prevent overfitting, generating an iterative strategy for the educational logic structure to ensure that the optimization of the educational logic structure conforms to the data change trend.

[0087] Further, as an embodiment of the present invention, refer to Figure 2 shown in Figure 1 the detailed step flow diagram of step S2 in

[0088] Further, step S2 includes the following steps:

[0089] Step S21: Conduct an educational structured feature analysis on the global mapping incremental educational matrix to generate educational structured feature data;

[0090] In the embodiments of the present invention, for the educational structured feature analysis of the global mapping incremental educational matrix, various structured data are first extracted from the global mapping incremental educational matrix, which usually includes students' basic information (such as grade, class, course selection, etc.), exam scores, attendance rate, homework completion status, etc. These data constitute the structured features in the educational process. For feature analysis, the principal component analysis (PCA) algorithm is used to perform dimensionality reduction on the extracted educational data to facilitate the identification of the main influencing factors and their relative weights. Through this step, it can be identified which educational features (such as course grades, homework completion degree, etc.) are most relevant to key indicators such as students' learning effects and participation. For example, the number of dimensions after PCA dimensionality reduction is set to 5 to ensure that at least 85% of the data variance can be retained. In addition, visualization tools (such as heat maps, scatter plots, etc.) are used to show the correlation between educational features, so as to further analyze the distribution and influence of structured features, thereby providing a basis for subsequent educational data analysis and decision-making.

[0091] Step S22: Perform educational unstructured key behavior event analysis based on the global mapping incremental educational matrix to obtain educational unstructured key behavior event data;

[0092] In the embodiments of the present invention, for the educational unstructured key behavior event analysis based on the global mapping incremental educational matrix, unstructured data are first identified from the incremental educational data, such as students' classroom interaction behaviors, learning discussions on social media, learning duration in online learning platforms, etc. These unstructured data need to be extracted and classified through natural language processing (NLP) techniques or behavior analysis models. An event detection model (such as an LSTM model based on deep learning) is used to analyze the key behavior events of students during the learning process, such as not completing homework for a long time or high-frequency interactive learning group discussions. By setting the detection threshold of the behavior event, if a student fails to complete two consecutive assignments in a learning module, the system regards this as a low-performance behavior and marks this behavior. The educational unstructured key behavior event data will provide a more targeted intervention signal for subsequent educational decision-making. For example, the recognition accuracy threshold of the unstructured key behavior event is set to 85% to ensure that the extracted key events can effectively reflect potential problems in students' learning and provide real-time data support for teaching intervention.

[0093] Step S23: Design educational unstructured feature mapping rules based on the educational unstructured key behavior event data, and use the educational unstructured feature mapping rules to perform educational unstructured feature mapping processing on the global mapping incremental educational matrix to generate educational unstructured feature data;

[0094] In the embodiments of the present invention, based on the educational unstructured key behavior event data, educational unstructured feature mapping rules are designed. First, the identified unstructured key behavior events (such as low-performance students, low course participation, low academic interaction frequency, etc.) need to be classified and labeled. Using classification algorithms (such as support vector machine SVM or decision tree algorithm) to conduct further feature analysis on educational unstructured behavior events, and extract key factors affecting students' performance, such as the frequency of homework submission, participation in online discussions, interaction frequency with teachers, etc. Then, design feature mapping rules to map these unstructured behavior features to the global mapping incremental education matrix. For example, when the frequency of a student's homework submission within a certain stage is lower than the set threshold (such as a homework on-time submission rate lower than 70%), then through the mapping rules, the behavior feature of this student can be marked as "low participation". In addition, when designing the rules, consider the consistency of educational background data to ensure that the mapping rules can adapt to data changes in different educational scenarios. The settings of the feature mapping rules include: frequency threshold (such as a homework submission rate lower than 70%); behavior frequency analysis window (such as analyzing based on weekly behavior data); mapping target feature dimension (such as "participation", "learning progress", etc.). Finally, apply the mapping rules to the global mapping incremental education matrix to generate educational unstructured feature data.

[0095] Step S24: Perform educational feature integration processing on the educational structured feature data and the educational unstructured feature data to generate educational feature data.

[0096] In the embodiments of the present invention, perform educational feature integration processing on the educational structured feature data and the educational unstructured feature data, and fuse the extracted educational structured feature data (such as students' exam scores, attendance, homework completion rate, etc.) with the educational unstructured feature data (such as students' classroom interaction frequency, online learning behavior, participation in social media discussions, etc.). Use machine learning models such as weighted average method, principal component analysis (PCA), or multi-layer perceptron (MLP) to conduct feature integration, and merge the two types of feature data according to certain weights to obtain comprehensive educational feature data. For example, adopt the PCA algorithm, set the variance ratio to be retained at 95% to ensure the diversity and representativeness of the feature data, and set the weights of the structured feature data and the unstructured feature data to be 0.7 and 0.3 respectively through the weighted average method for data integration. The integrated educational feature data will cover multiple dimensions such as students' academic performance, behavior patterns, learning attitudes, etc., as well as the corresponding weight information, forming a complete educational feature dataset. The result of this step can provide support for subsequent educational decisions. For example, it can help teachers design personalized teaching strategies or optimize learning path recommendations, etc.

[0097] Furthermore, step S22 includes the following steps:

[0098] Step S221: Conduct educational activity behavior analysis based on the global mapping incremental education matrix to generate educational activity behavior data; and perform educational activity behavior object detection through the educational activity behavior data to generate educational activity behavior object data;

[0099] Step S222: Use the educational activity behavior data to perform educational unstructured behavior parsing on the global mapping incremental education matrix to generate educational unstructured behavior parsing data;

[0100] Step S223: Conduct educational unstructured behavior pattern analysis based on the educational activity behavior object data and the educational unstructured behavior parsing data to generate educational unstructured behavior pattern data;

[0101] Step S224: Conduct educational unstructured key behavior event analysis on the educational unstructured behavior parsing data according to the educational unstructured behavior pattern data to obtain educational unstructured key behavior event data.

[0102] In the embodiments of the present invention, educational activity behaviors are analyzed based on the global mapping incremental education matrix. This step can be carried out through time series analysis or data clustering methods, with the aim of extracting the behavior patterns of students in educational activities. For example, behavioral data such as the click frequency of students during the learning process, the submission of assignments, and the completion progress of learning modules can be extracted through time series analysis methods to generate educational activity behavior data. The clustering algorithm (such as K-means) is used to classify students' behaviors to discover groups of students under different learning modes. Then, by performing behavior object detection on the educational activity behavior data, specific educational activity behavior objects are identified, such as a certain student's participation in a course, the frequency of assignment submission, and online discussion interactions. This is achieved through object detection algorithms (such as YOLO or Faster R-CNN) to identify and label the learning activity behaviors of students. For example, a threshold is set (such as an assignment submission rate below 70% being "low participation") to identify behavior objects, generating educational activity behavior object data. The educational activity behavior data is used to perform educational unstructured behavior analysis on the global mapping incremental education matrix, and unstructured behavior data of students during the learning process is extracted from the global mapping incremental education matrix. For example, students' emotional states, the frequency of interaction with teachers, and the number of times of participating in classroom discussions can be analyzed through sentiment analysis techniques on students' speech or text data (such as comments and questions on the online learning platform) or through sensor data (such as learning progress and the frequency of touch screen operations) for behavior analysis. Commonly used sentiment analysis algorithms include sentiment classification models in natural language processing, and through these technologies, students' emotional states (such as anxiety, confusion, positive, etc.) are identified. Based on these analysis results, students' learning status data can be further obtained, such as whether they are in an actively participating or confused state. Through educational unstructured behavior analysis, educational unstructured behavior analysis data is generated, and these data will help to analyze students' behaviors in more detail, and then optimize educational intervention strategies. Educational unstructured behavior pattern analysis is carried out based on educational activity behavior object data and educational unstructured behavior analysis data. This process mainly identifies and analyzes the unstructured behavior patterns of students in educational activities, such as learning preferences, learning rhythms, and emotional changes. First, educational activity behavior object data is collected, including information such as students' learning participation, classroom interaction data, and online activities. At the same time, combined with the analysis results of unstructured data such as sentiment analysis, data such as students' learning status and emotional fluctuations are obtained. Then, clustering analysis or unsupervised learning methods in deep learning are applied to perform pattern recognition on these multi-dimensional data to find similar behavior patterns. For example, a group of students may show high participation for several consecutive days, while another group of students may have intermittent low participation and show emotional fluctuations. These different behavior patterns will be classified and labeled, and finally educational unstructured behavior pattern data is generated.By setting a certain clustering threshold (such as behavior frequency, emotional fluctuation range, etc.) and combining with the output of the model, different learning patterns of the student group are identified. In this way, the behavior patterns of students are identified, providing data support for subsequent educational interventions and personalized recommendations. Based on the educational unstructured behavior pattern data, educational unstructured key behavior event analysis is carried out on the educational unstructured behavior parsing data, and key behavior events are extracted from the educational unstructured behavior pattern data. For example, the emotional changes of students are too large, and the learning activities suddenly stagnate, etc. First, through the pattern data obtained from the analysis of educational unstructured behavior patterns, potential key events are identified, such as some students showing significant emotional fluctuations or drastic changes in learning status at a certain learning stage. Then, according to the set event thresholds (such as emotional fluctuations exceeding a certain value, the time period when learning behaviors change, etc.), event detection is carried out on the educational unstructured behavior parsing data. For example, when the emotional fluctuation exceeds a certain set threshold, it can be determined as an "emotional fluctuation event"; when the learning progress suddenly drops or stagnates, it can be regarded as a "learning stagnation event". Technologies such as time series analysis (such as ARIMA) and deep learning (such as LSTM) are used to analyze the temporal changes of students' behavior patterns and detect the occurrence of key events. Finally, the analysis results are aggregated to obtain educational unstructured key behavior event data. These key event data can help educators timely identify possible learning problems or emotional problems of students and provide support for formulating personalized intervention strategies and educational suggestions.

[0103] Further, step S224 includes the following steps:

[0104] Design an educational key behavior event detection threshold through educational unstructured behavior pattern data;

[0105] Conduct temporal analysis of behavior patterns based on educational unstructured behavior pattern data, generate educational unstructured behavior pattern temporal data, and design an educational key behavior event temporal detection threshold through the educational unstructured behavior pattern temporal data;

[0106] Extract educational unstructured key behavior events from the educational unstructured behavior parsing data that meet the educational key behavior event detection threshold and the educational key behavior event temporal detection threshold to obtain educational unstructured key behavior event data.

[0107] In the embodiments of the present invention, by analyzing the unstructured educational behavior pattern data, appropriate detection thresholds for key educational behavior events are designed. First, according to the behavior pattern data of students (such as learning progress, emotional fluctuations, participation, etc.), statistical analysis methods (such as mean, standard deviation, quantiles, etc.) are used to set the thresholds. For example, in the case of emotional fluctuations, it is set that when the change in the student's emotional index is greater than twice the standard deviation, it is regarded as a large emotional fluctuation and may require attention. For the change in learning progress, if the student's learning progress shows a significant decline compared to their normal trajectory, a threshold can be set to identify potential stagnation phenomena (such as a reduction in learning time by more than 30%). By adjusting these thresholds, it is ensured that key behavior events can be detected in a timely manner when the data changes greatly or abnormally, providing support for subsequent intelligent teaching interventions. Perform time series analysis on the unstructured educational behavior pattern data to generate unstructured educational behavior pattern time series data. First, model the behavior data through time series analysis (such as ARIMA, LSTM and other models) to extract the trends, seasonal fluctuations and periodic changes of the data. For example, if the student's learning participation shows periodic fluctuations, then the learning patterns shown during specific time periods (such as before and after exams, holidays) can be identified. Through this analysis, unstructured educational behavior pattern time series data can be generated, including the learning status, emotional fluctuations, etc. of each student at different time periods. Subsequently, design time series detection thresholds for key educational behavior events, and these thresholds will be set based on the behavior pattern time series data. For example, if a student has emotional fluctuations exceeding 2 standard deviations within a week and lasts for more than the set time, it can be set as a "key behavior event", and this threshold setting is used for subsequent key event identification. By meeting the conditions of the detection threshold for key educational behavior events and the time series detection threshold for key educational behavior events, eligible unstructured key educational behavior events are screened out. Extract the learning behaviors of each student through the unstructured educational behavior analysis data, such as learning participation, emotional fluctuations, interaction frequency, etc., and compare them with the set detection thresholds one by one. If a student's behavior meets the above threshold conditions and shows abnormalities or continuous changes (such as continuous emotional fluctuations) in the time series analysis, it is determined as a key behavior event. Through these data screenings, unstructured key educational behavior event data can be extracted. For example, if a student shows emotional fluctuations exceeding the preset threshold and lasting for more than the set time during the learning period, and at the same time their learning activities stop, it will be determined as a "learning stagnation" event.

[0108] Further, as an embodiment of the present invention, refer to Figure 3 shown, for Figure 1 the detailed step flow diagram of step S3 in

[0109] Step S31: Perform clustering processing on educational feature data in terms of educational feature dimensions to generate educational feature dimension clustering data;

[0110] In the embodiments of the present invention, clustering analysis is performed on educational feature data with the aim of discovering potential patterns in the data and grouping relevant features, thereby simplifying the subsequent evaluation and analysis process. The K-means clustering algorithm is used for clustering educational feature dimensions. The specific parameter settings of the K-means algorithm are as follows: The value of K is selected as 3, indicating that the data is divided into three categories; the initial centroids are randomly selected, the number of clustering iterations is set to 50 times, and the clustering accuracy is controlled within 0.001. Through this clustering process, educational feature dimension clustering data is generated, where each clustering result corresponds to a group of students with similar educational features. These clustering results help identify the behavior patterns of different student groups.

[0111] Step S32: Perform classification processing on educational evaluation indicators based on the educational feature dimension clustering data to generate educational evaluation classification indicator data;

[0112] In the embodiments of the present invention, classification processing of educational evaluation indicators is performed based on the educational feature dimension clustering data. According to the clustering results of different feature dimensions, the learning behaviors and performances of each category of students are analyzed and mapped into a preset educational evaluation indicator system. For example, for the high-achievement group, indicators such as "academic performance" and "exam passing rate" are mainly considered; while for the low-participation group, more attention is paid to behavioral indicators such as "learning duration" and "interaction frequency". Classification of educational evaluation indicators can be carried out through methods such as hierarchical clustering or decision trees. Using a decision tree model, classification indicators are generated based on the clustering results of students and various feature data. The set classification basis includes key dimensions such as students' participation, academic performance, and learning behaviors. Through this process, educational evaluation classification indicator data is generated, providing structured data support for the subsequent evaluation model.

[0113] Step S33: Design educational evaluation category indicators and corresponding educational evaluation category subset indicators based on the educational evaluation classification indicator data;

[0114] In the embodiments of the present invention, based on the educational evaluation classification index data, educational evaluation category indicators and corresponding subset indicators are further designed. According to the characteristics of each group, suitable evaluation categories are designed. For example, for the academic performance group, the "academic ability" category is designed, and the following subset indicators can include "exam scores", "assignment completion", "academic contribution", etc.; for the behavioral performance group, the "learning attitude" category is designed, and the subset indicators can include "classroom participation", "timely assignment submission rate", "after-class feedback", etc. The weights of each category indicator and subset indicator are determined through expert review combined with statistical analysis methods. For example, the expert scoring method combined with the AHP (Analytic Hierarchy Process) model is used for the weight allocation of evaluation categories to ensure the rationality and comparability of each category indicator, so as to generate educational evaluation category indicators and subset indicators, providing a basis for educational level evaluation and subsequent intelligent evaluation.

[0115] Step S34: Obtain historical educational evaluation data;

[0116] In the embodiments of the present invention, historical educational evaluation data is obtained from educational data sources or systems. The historical educational evaluation data can include students' grade records, classroom performance data, assignment feedback, educational evaluation records, etc. Relevant data is extracted from the database through the API interface of the educational administration system.

[0117] Step S35: Conduct an analysis of the multi-level quantification evaluation characteristics of educational evaluation category subset indicators based on the historical educational evaluation data to generate multi-level quantification evaluation characteristic data of educational indicators;

[0118] In the embodiments of the present invention, a multi-level scoring model (such as the Analytic Hierarchy Process AHP) is used to conduct multi-level quantification evaluation of each subset indicator. For example, for the academic ability category of students, it can be quantitatively analyzed from three dimensions: "exam scores", "assignment completion", and "classroom performance". By using a multi-level scoring system, the evaluation of each dimension is divided into several grades. For example, the historical educational data is standardized so that different types of data can be compared uniformly. For example, the exam scores are standardized so that the scores of different subjects have the same quantification standard. According to the nature of each subset indicator, quantification criteria are set. The Analytic Hierarchy Process (AHP) is combined with expert review to weight the relative importance of each subset indicator to obtain the weight of each indicator in the overall score. In this way, a quantification evaluation model for each subset indicator is constructed. Combining the historical educational evaluation data and the quantification criteria, an analysis of the multi-level quantification characteristics of each educational evaluation category subset indicator is conducted to generate multi-level quantification evaluation characteristic data. For example, analyzing the academic performance of students in a certain semester, the quantification evaluation data generated based on exam scores, assignment completion, and classroom performance can include the quantification scores and comprehensive evaluation scores of each subset under the "academic performance" category.

[0119] Step S36: Integrate and process the educational level evaluation indicators by using the corresponding educational evaluation category indicators, educational evaluation category subset indicators, and multi-level quantitative evaluation characteristic data of educational indicators to generate educational level evaluation indicator data;

[0120] In the embodiment of the present invention, through the comprehensive analysis of the educational evaluation category indicators, educational evaluation category subset indicators, and multi-level quantitative evaluation characteristic data of educational indicators, the integration and processing of the educational level evaluation indicators are carried out, and finally the educational level evaluation indicator data is generated. Technical means such as the weighted average method, hierarchical weighted method, or factor analysis method are used to integrate the evaluation data of each category indicator and subset indicator to generate a comprehensive evaluation indicator. For example, for the comprehensive evaluation of students, multiple subset indicators (such as exam scores, homework completion, and class participation) under the category of "academic performance" can be weighted according to the preset weights to generate the comprehensive evaluation data of the category of "academic performance". Collect the quantitative data of the educational evaluation category indicators, educational evaluation category subset indicators, and educational indicators. These data can be sourced from historical educational evaluation data or obtained through a real-time data acquisition system. According to the relative importance of the educational evaluation categories and subset indicators, the weight of each indicator is determined through expert review or statistical methods (such as the Analytic Hierarchy Process AHP). The quantitative data of each subset indicator is weighted and calculated according to the weights to obtain the comprehensive evaluation data of each educational category (such as academic performance, social practice, etc.), and finally the overall educational level evaluation indicator data is formed.

[0121] Step S37: Design the tree-level mapping relationship between educational activities and educational evaluations based on the educational level evaluation indicator data to obtain the intelligent education evaluation level relationship model.

[0122] In the embodiments of the present invention, based on the educational level evaluation index data, a tree - shaped hierarchical mapping relationship between educational activities and educational evaluations is designed to form a hierarchical relationship model of an intelligent education evaluation system (intelligent education evaluation). This model is used to automatically analyze all aspects of educational activities and optimize and adjust according to the evaluation data at each level. Design a tree - shaped structure, decompose different levels in the educational evaluation process into different nodes, and map the evaluation data of each node accordingly according to the educational level evaluation index data. According to the educational level evaluation index data, first design the hierarchical relationship tree of educational activities and evaluations. For example, the root node of the educational level tree can be "Comprehensive Educational Evaluation", which is divided into several first - level nodes such as "Academic Ability", "Innovation Ability", "Social Practice", etc., that is, the educational evaluation category indicators. Each first - level node is further divided into multiple second - level nodes (for example, under "Academic Ability", there are educational evaluation category subset indicators such as "Exam Scores", "Classroom Performance", "Homework Completion Degree", etc.). According to the evaluation data of each level, map it to the corresponding educational activities. Based on the designed tree - shaped structure, an automatic hierarchical mapping relationship model between educational activities and evaluations is constructed through intelligent algorithms (such as the Analytic Hierarchy Process, neural network models, etc.). Through the calculation of the model, the evaluation weights and mapping rules of each level can be automatically adjusted to optimize the implementation strategy of educational activities. Verify the initially designed tree - shaped hierarchical relationship model according to historical data, and evaluate its prediction accuracy and adaptability. If the model performs poorly, adjust the weights and mapping rules to further optimize the model.

[0123] Further, step S33 includes the following steps:

[0124] Step S331: Design educational evaluation category indicators through educational evaluation classification index data;

[0125] Step S332: Analyze the factor influencing factors of classification indicators according to educational evaluation classification index data to generate classification indicator factor influencing factor data;

[0126] Step S333: Design educational evaluation category subset indicators through the classification indicator factor influencing factor data corresponding to educational evaluation category indicators.

[0127] In the embodiments of the present invention, appropriate educational evaluation category indicators are designed through educational evaluation classification index data. These category indicators will provide a basis for the subsequent educational evaluation system and help further divide and analyze specific educational evaluation dimensions. Using the hierarchical design method, different category indicators are designed according to different dimensions of the educational evaluation classification index. These category indicators should be able to comprehensively reflect all aspects of educational activities, such as students' academic performance, innovation ability, social practice, etc. For example, historical educational evaluation data, teaching activity data, etc. are collected and sorted out, and relevant educational evaluation classification index data is extracted. Based on the content and characteristics of the classification index, the categories of educational evaluation are determined. According to the educational evaluation classification index data, each sub-category is integrated to form the corresponding category indicator. Factor analysis is carried out using the educational evaluation classification index data to identify the influencing factors for each classification index, and classification index factor influencing factor data is generated to understand the formation reasons and influencing factors of each educational evaluation index, so as to provide data support for subsequent optimization and adjustment. Statistical methods such as factor analysis method and regression analysis method are used to analyze the influencing factors of the classification index. For example, data on educational evaluation classification indexes, such as students' academic achievements, classroom behaviors, participation degrees, etc. are collected, and the relevant factors of these indexes are further recorded. Using the factor analysis method or the regression analysis method, the factor influence of each educational evaluation classification index is analyzed. Through the obtained influencing factors, factor influencing factor data is generated to further provide a basis for the design of educational evaluation category indicators. Using the factor influencing factor data of the classification index, educational evaluation category subset indicators are further designed. These subset indicators will provide support for the refinement and precision of the educational evaluation system and help improve the accuracy and practicality of educational evaluation. According to the identified factor influencing factors, subset indicators are set and mapped to the corresponding educational evaluation categories. For example, according to the analysis of the factor influencing factor data of the classification index, the specific factors affecting each classification index are analyzed. According to these factors, relevant subset indicators are designed to further refine the educational evaluation system and ensure that the evaluation results can accurately reflect the effects of educational activities.

[0128] Further, step S4 includes the following steps:

[0129] Step S41: Based on the global educational data and the instant incremental educational data, perform an educational evaluation requirement task analysis on the educational feature data to generate educational evaluation requirement task data;

[0130] Step S42: According to the educational evaluation requirement task data, perform a model dynamic adjustment and optimization process on the intelligent educational evaluation hierarchical relationship model for the educational evaluation requirement task to generate an optimized intelligent educational evaluation hierarchical relationship model;

[0131] Step S43: Transmit the educational feature data to the optimized intelligent educational evaluation hierarchical relationship model for educational hierarchical intelligent evaluation processing to generate educational hierarchical intelligent evaluation data;

[0132] Step S44: Conduct an educational intelligent evaluation collaboration analysis based on the educational level intelligent evaluation data, generate educational intelligent evaluation data, and transmit the educational intelligent evaluation data to the terminal to execute the intelligent feedback operation of educational evaluation.

[0133] In the embodiments of the present invention, through the comprehensive analysis of global education data and instant incremental education data, the demand tasks of education feature data are analyzed to identify the key demands and tasks existing in the education evaluation process. This analysis can provide data support for the design and implementation of education evaluation tasks and help formulate more accurate evaluation strategies. The goal of task analysis is to clarify which aspects of education features need to be concerned and optimized to ensure the effectiveness and pertinence of education evaluation. Global education data and instant incremental education data are collected, and the data sources include student achievement data, learning behavior data, teacher feedback data, student participation data, etc. Global education data usually includes historical education data, while instant incremental education data includes recently generated data, which helps to timely reflect the changes in the education process, extract the features required for education evaluation, and identify the key demands in the education process. According to the analysis results of education feature data, education evaluation demand task data is generated. The demand task data includes task objectives, task priorities, task execution times, etc. According to the education evaluation demand task data, the intelligent education evaluation hierarchical relationship model is dynamically adjusted and optimized. The intelligent education evaluation hierarchical relationship model analyzes and evaluates according to different dimensions of education data through a multi-level evaluation mechanism. Through dynamic adjustment and optimization, it can ensure that the model can adapt to the demand changes in different education scenarios in real time, improving the flexibility and response ability of the model. The education evaluation demand task data is input into the intelligent education evaluation hierarchical relationship model. This data includes the priority, resource requirements, task objectives, etc. of the task. The model needs to make corresponding adjustments according to these input data to meet specific education evaluation requirements. Technologies such as reinforcement learning and adaptive optimization algorithms are used to adjust and optimize the intelligent education evaluation hierarchical relationship model. For example, the particle swarm optimization (PSO) algorithm or genetic algorithm (GA) is used to optimize the model. By adjusting parameters such as the evaluation weights and processing methods of the model, it can adapt to specific education evaluation tasks. The model optimization process is dynamically adjusted according to factors such as task priority, time limit, and data consistency. The dynamic adjustment of the intelligent education evaluation hierarchical relationship model generates an optimized model. The optimized model can respond more accurately to education evaluation requirements and has stronger adaptability. For example, in practical applications, some tasks may require higher precision and efficiency, while other tasks can be carried out under slightly looser conditions. The optimized model ensures the accuracy and timeliness of education evaluation. The optimized model has been dynamically adjusted according to the education evaluation demand task data to improve the adaptability and accuracy of the model in various education evaluation tasks. The education feature data that has undergone education evaluation demand task analysis is transmitted to the optimized intelligent education evaluation hierarchical relationship model. The education feature data usually includes students' learning behavior data, learning paths, learning achievements, participation in teaching activities, etc. These data reflect the performance of students in the education process and the implementation of education activities.Using the intelligent algorithms in the model, such as evaluating and defining educational evaluation through fuzzy logic and conducting learning evaluation through deep learning, multi-level intelligent evaluation processing is performed on the input educational feature data. The model will evaluate educational features based on different levels of evaluation criteria. For example, a student's learning progress, academic performance, and participation in a certain course module may be assigned different weights and scoring rules according to a predefined hierarchical structure. According to the intelligent processing results, educational hierarchical intelligent evaluation data is generated. This data includes evaluation indicators at multiple levels, such as a student's comprehensive learning ability, learning interest, participation, etc., and is accompanied by corresponding recommended measures and optimization suggestions. For example, if the model finds that a student has a low academic performance in a specific area, the model may recommend personalized learning resources or adjust the learning path. According to the educational hierarchical intelligent evaluation data, collaborative analysis algorithms (such as collaborative filtering, ensemble learning, etc.) are used to comprehensively analyze the evaluation data at different levels. The collaborative analysis method can obtain a more comprehensive and accurate evaluation result by integrating multiple evaluation dimensions. In this way, the model can discover potential relationships and patterns and provide more personalized feedback. According to the results of the collaborative analysis, educational intelligent evaluation data is generated. These data may include optimized teaching strategies, personalized learning path adjustments, teaching resource allocation suggestions, etc. The generated feedback data can help teachers and educational administrators make more accurate decisions. For example, if a student performs poorly in a certain course, it is recommended to adjust the teaching strategy for that course or provide the student with more learning resources. The generated educational intelligent evaluation data is transmitted to terminal devices, such as teacher terminals, student terminals, etc. These terminals will execute subsequent intelligent feedback operations for educational evaluation to ensure that the feedback can be transmitted to relevant personnel in a timely and effective manner, and educational evaluation and feedback can be automated and precise, improving the educational effect.

[0134] This specification provides a multi-modal data-driven educational evaluation system for performing the multi-modal data-driven educational evaluation method as described above. The multi-modal data-driven educational evaluation system includes:

[0135] An educational data collection module, which is used to collect multi-modal educational data by using a preset educational administration system API interface, generate multi-modal educational data; perform global integration and incremental iterative mapping processing of educational data based on the multi-modal educational data, and generate a global mapping incremental educational matrix;

[0136] An educational feature analysis module, which is used to perform structured and unstructured educational feature analysis based on the global mapping incremental educational matrix and generate educational feature data;

[0137] The teaching evaluation hierarchical relationship model design module is used to obtain historical education evaluation data; analyze the education hierarchical evaluation indicators based on the education characteristic data and the historical education evaluation data to generate education hierarchical evaluation index data; design the tree-shaped hierarchical mapping relationship between educational activities and educational evaluations based on the education hierarchical evaluation index data to obtain the intelligent teaching evaluation hierarchical relationship model;

[0138] The intelligent teaching evaluation feedback module is used to perform intelligent education evaluation processing on the education characteristic data based on the intelligent teaching evaluation hierarchical relationship model, generate intelligent education evaluation data, and transmit the intelligent education evaluation data to the terminal to execute the intelligent feedback operation of teaching evaluation.

[0139] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be encompassed within the present invention.

[0140] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A multi-modal data-driven education evaluation method, characterized in that It includes the following steps: Step S1: Use a preset educational administration system API interface to collect multi-modal educational data and generate multi-modal educational data; perform global integration and incremental iterative mapping processing on the educational data based on the multi-modal educational data to generate a global mapping incremental educational matrix; Step S2: Perform structured and unstructured educational feature analysis based on the global mapping incremental educational matrix to generate educational feature data; Among them, Step S2 includes: Step S21: Perform educational structured feature analysis on the global mapping incremental educational matrix to generate educational structured feature data; Step S22: Perform educational unstructured key behavior event analysis based on the global mapping incremental educational matrix to obtain educational unstructured key behavior event data; Step S22 includes: Step S221: Perform educational activity behavior analysis based on the global mapping incremental educational matrix to generate educational activity behavior data; and perform educational activity behavior object detection through the educational activity behavior data to generate educational activity behavior object data; Step S222: Use the educational activity behavior data to perform educational unstructured behavior parsing on the global mapping incremental educational matrix to generate educational unstructured behavior parsing data; Step S223: Perform educational unstructured behavior pattern analysis based on the educational activity behavior object data and the educational unstructured behavior parsing data to generate educational unstructured behavior pattern data; Step S224: Design an educational key behavior event detection threshold through the educational unstructured behavior pattern data; perform behavior pattern time series analysis based on the educational unstructured behavior pattern data to generate educational unstructured behavior pattern time series data, and design an educational key behavior event time series detection threshold through the educational unstructured behavior pattern time series data; extract the educational unstructured key behavior event corresponding to the educational key behavior event detection threshold and the educational key behavior event time series detection threshold from the educational unstructured behavior parsing data to obtain educational unstructured key behavior event data; Step S23: Design an educational unstructured feature mapping rule based on the educational unstructured key behavior event data, and use the educational unstructured feature mapping rule to perform educational unstructured feature mapping processing on the global mapping incremental educational matrix to generate educational unstructured feature data; Step S24: Perform educational feature integration processing on the educational structured feature data and the educational unstructured feature data to generate educational feature data; Step S3: Obtain historical educational evaluation data; perform educational level evaluation index analysis based on the educational feature data and the historical educational evaluation data to generate educational level evaluation index data; design a tree-level mapping relationship between educational activities and educational evaluations based on the educational level evaluation index data to obtain an intelligent educational evaluation level relationship model; Step S4: Perform educational intelligent evaluation processing on the educational feature data based on the intelligent educational evaluation level relationship model to generate educational intelligent evaluation data, and transmit the educational intelligent evaluation data to the terminal to execute the intelligent feedback operation of educational evaluation.

2. The multi-modal data-driven education evaluation method according to claim 1, wherein Step S1 includes the following steps: Step S11: Use the preset educational administration system API interface to collect multi-modal education data and generate multi-modal education data; Step S12: Design an educational logic structure relationship matrix based on the multi-modal education data and generate an educational logic structure relationship matrix; Step S13: Design an educational logic structure iteration strategy based on the educational logic structure relationship matrix and generate an educational logic structure iteration strategy; Step S14: Collect global education data according to the multi-modal education data, generate global education data, and transmit the global education data to the educational logic structure relationship matrix for preliminary mapping processing of educational data to generate a global mapping education matrix; Step S15: Instantly update the incremental education data according to the multi-modal education data to generate instant incremental education data, and transmit the instant incremental education data to the global mapping education matrix through the educational logic structure iteration strategy for iterative mapping processing of incremental education data to generate a global mapping incremental education matrix.

3. The multimodal data-driven education evaluation method according to claim 2, wherein Step S13 includes the following steps: Step S131: Analyze the educational logic structure features based on the educational logic structure relationship matrix and generate educational logic structure feature data; Step S132: Analyze the educational logic structure iteration requirements based on the educational logic structure feature data and generate educational logic structure iteration requirement data; Step S133: Analyze the educational data standardization requirements according to the educational logic structure feature data and the data types of the multi-modal education data, and generate educational data standardization requirement data; Step S134: Design an educational logic structure iteration strategy through the educational logic structure iteration requirement data and the educational data standardization requirement data, and generate an educational logic structure iteration strategy.

4. The multi-modal data-driven education evaluation method according to claim 1, characterized in that Step S3 includes the following steps: Step S31: Perform clustering processing on the educational feature dimensions of the educational feature data to generate educational feature dimension clustering data; Step S32: Classify the educational evaluation indicators according to the educational feature dimension clustering data to generate educational evaluation classification index data; Step S33: Design educational evaluation category indicators and corresponding educational evaluation category subset indicators based on the educational evaluation classification index data; Step S34: Obtain historical educational evaluation data; Step S35: Analyze the multi-level quantization evaluation characteristics of the educational evaluation category subset indicators according to the historical educational evaluation data to generate educational index multi-level quantization evaluation characteristic data; Step S36: Integrate the educational evaluation category indicators, the educational evaluation category subset indicators, and the educational index multi-level quantization evaluation characteristic data to generate educational level evaluation index data; Step S37: Design a tree-level mapping relationship between educational activities and educational evaluations based on the educational level evaluation index data to obtain an intelligent teaching evaluation level relationship model.

5. The multimodal data-driven education evaluation method according to claim 4, wherein Step S33 includes the following steps: Step S331: Design educational evaluation category indicators through the educational evaluation classification index data; Step S332: Analyze the factor influencing factors of the classification indicators according to the educational evaluation classification index data to generate classification indicator factor influencing factor data; Step S333: Design the subset indicators of the education evaluation category by using the influencing factor data of the classification index factors corresponding to the education evaluation category indicators.

6. The multimodal data-driven education evaluation method according to claim 1, characterized in that Step S4 includes the following steps: Step S41: Conduct an analysis of the education evaluation requirement tasks on the education feature data based on the global education data and the instant incremental education data, and generate the education evaluation requirement task data. Step S42: Perform model dynamic adjustment and optimization processing of the education evaluation requirement tasks on the intelligent education evaluation hierarchical relationship model according to the education evaluation requirement task data, and generate the optimized intelligent education evaluation hierarchical relationship model. Step S43: Transmit the education feature data to the optimized intelligent education evaluation hierarchical relationship model for intelligent evaluation processing of the education level, and generate the intelligent evaluation data of the education level. Step S44: Conduct an analysis of the education intelligent evaluation collaboration based on the intelligent evaluation data of the education level, generate the education intelligent evaluation data, and transmit the education intelligent evaluation data to the terminal to execute the intelligent feedback operation of the education evaluation.

7. A multimodal data-driven education evaluation system, characterized in that, For implementing the multi-modal data-driven education evaluation method as described in Claim 1, the multi-modal data-driven education evaluation system includes: An education data collection module, which is used to collect multi-modal education data by using the preset API interface of the educational administration system, and generate multi-modal education data; perform global integration and incremental iterative mapping processing of the education data based on the multi-modal education data, and generate the global mapping incremental education matrix. An education feature analysis module, which is used to conduct structured and unstructured education feature analysis based on the global mapping incremental education matrix, and generate education feature data. An education evaluation hierarchical relationship model design module, which is used to obtain historical education evaluation data; conduct an analysis of the education level evaluation indicators according to the education feature data and the historical education evaluation data, and generate the education level evaluation index data; design the tree-shaped hierarchical mapping relationship between the education activities and the education evaluation based on the education level evaluation index data to obtain the intelligent education evaluation hierarchical relationship model. An intelligent education evaluation feedback module, which is used to conduct intelligent education evaluation processing on the education feature data based on the intelligent education evaluation hierarchical relationship model, generate the education intelligent evaluation data, and transmit the education intelligent evaluation data to the terminal to execute the intelligent feedback operation of the education evaluation.

Images