Education evaluation auxiliary method and system based on artificial intelligence

Abstract

The invention discloses an education evaluation assisting method and system based on artificial intelligence. The method comprises the steps of data collection, data optimization, learning behavior evaluation, education feature evaluation and education evaluation assisting. The invention relates to the technical field of education management, in particular to an education evaluation auxiliary method and system based on artificial intelligence. According to the scheme, a two-way correlation analysis method combining learning behaviors and education features is adopted, preliminary analysis is conducted through learning behavior evaluation, multi-dimensional extraction is conducted through the education features, and the education evaluation accuracy is improved. The analysis depth and the analysis specificity of education evaluation are improved, and meanwhile, an effective way for assisting the education evaluation by an intelligent method is provided; performing learning behavior evaluation by adopting a method of combining a causal enlightenment enhanced graph nerve long-short-term memory integration model; and carrying out educational feature evaluation by adopting a graph neural network method combining knowledge migration and a knowledge graph.

Description

Technical Field

[0001] The present invention relates to the field of education management technology, and specifically to an education assessment auxiliary method and system based on artificial intelligence. Background Art

[0002] The AI-based auxiliary method for educational assessment is an intelligent method that uses AI technology to conduct a comprehensive analysis of students' learning behaviors, learning outcomes, and psychological states. It collects and processes multimodal data and combines algorithms to build personalized assessment models to accurately diagnose students' learning difficulties and ability characteristics. The system can not only provide real-time personalized feedback and learning suggestions, but also assist teachers in optimizing teaching strategies, thereby improving education quality and learning efficiency. Its core role is to realize the intelligence, precision, and personalization of educational assessment, and provide a scientific basis for teaching students in accordance with their aptitude.

[0003] However, among the existing auxiliary methods for educational assessment, there are problems with the existing intelligent solutions for educational assessment, such as data uniformity and single assessment dimension, which leads to the technical problem that the assessment results can neither properly reflect the actual educational status of students nor be effectively intelligent; among the existing learning behavior assessment methods, there is the problem that learning behavior often involves complex and deep high- and low-dimensional time-space characteristics, and traditional methods have limited analysis capabilities for multimodal basic data, which makes it difficult to effectively present the changes in learning behavior over time trends, and it is also impossible to intuitively reflect the results of learning behavior assessment, resulting in technical problems of low readability, low comprehensibility and low usability; among the existing educational feature assessment methods, there is the technical problem that educational features involve complex data from multiple aspects and angles, and existing technologies cannot effectively decompose and conduct comprehensive assessments. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an educational assessment auxiliary method and system based on artificial intelligence. In view of the fact that in the existing educational assessment auxiliary methods, there are problems that the existing educational assessment intelligent solutions often have single data and single assessment dimensions, which leads to the technical problem that the assessment results can neither properly reflect the actual educational status of students nor can they be effectively intelligent. This solution creatively adopts a two-way correlation analysis method combining learning behavior and educational characteristics, conducts preliminary analysis through learning behavior assessment, and refines educational characteristics in multiple dimensions, thereby improving the analysis depth and specificity of educational assessment, and at the same time provides an effective way for intelligent methods to assist educational assessment; in view of the fact that in the existing learning behavior assessment methods, learning behavior often involves complex and deep high and low latitude time and space characteristics, and traditional methods have limited analysis capabilities for multimodal basic data, which leads to the technical problem that it is difficult to effectively present the changes in learning behavior in time trends, and it is also impossible to intuitively reflect the results of learning behavior assessment, resulting in low readability, low comprehensibility and low usability. This solution creatively adopts a two-way correlation analysis method combining learning behavior and educational characteristics, conducts preliminary analysis through learning behavior assessment, and refines educational characteristics in multiple dimensions, thereby improving the analysis depth and specificity of educational assessment, and at the same time provides an effective way for intelligent methods to assist educational assessment; in view of the fact that in the existing learning behavior assessment methods, learning behavior often involves complex and deep high and low latitude time and space characteristics, and traditional methods have limited analysis capabilities for multimodal basic data, which leads to the technical problem that it is difficult to effectively present the changes in learning behavior in time trends, and it is also impossible to intuitively reflect the results of learning behavior assessment, resulting in low readability, low comprehensibility and low usability. The method of combining causal-inspired enhanced graph neural long-short-term memory integrated model is used to conduct learning behavior evaluation. The causal relationship between input data is enhanced by the causal-inspired algorithm, and the time-space characteristics are optimized by constructing an improved capsule network, which improves the prediction accuracy and efficiency of learning behavior evaluation as a basic task, improves the effectiveness of the method and provides a strong model and data foundation for subsequent steps. In view of the technical problem that in the existing educational feature evaluation methods, there are complex data involving multiple aspects and angles of educational features, and the existing technology cannot effectively decompose and conduct comprehensive evaluation, this solution creatively adopts a graph neural network method that combines knowledge transfer and knowledge graph. By constructing an educational feature knowledge graph embedded in the graph neural network, it can effectively supplement the educational feature analysis information of other dimensions required for learning behavior evaluation. Through transfer learning, it can not only effectively inherit the data results of learning behavior evaluation, but also improve the overall operation efficiency of the model, providing a good operation process for the comprehensive and complex task of assisting educational evaluation. At the same time, the intuitive quantitative data and clear classification data result output also effectively improve the reference and readability of the evaluation results.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides an educational assessment auxiliary method based on artificial intelligence, the method comprising the following steps:

[0006] Step S1: data collection;

[0007] Step S2: data optimization;

[0008] Step S3: learning behavior assessment;

[0009] Step S4: Educational characteristics assessment;

[0010] Step S5: Educational assessment assistance.

[0011] Furthermore, in step S1, the data collection is used to collect the original data set required for educational assessment, specifically, to obtain the original data set for educational assessment through learning scenario behavior data collection;

[0012] The learning scenario behavior data collection specifically includes classroom data collection, extracurricular activity data collection and homework data collection;

[0013] The original educational assessment data set specifically includes learning behavior data, emotional psychology data, and text record data;

[0014] The learning behavior data is specifically collected from the education management system data records, including classroom attention concentration assessment data, classroom discussion participation data and classroom discipline data;

[0015] The emotional and psychological data are specifically collected through psychological measurement questionnaires, including students' mental health data and students' emotional state record data;

[0016] The text record data is specifically collected from homework content and classroom speech records, including student composition homework record data and student classroom speech record data.

[0017] Further, in step S2, the data optimization is used to optimize the original data, specifically to enhance and optimize the original education assessment data set to obtain an education assessment optimized data set, and specifically includes the following steps:

[0018] Step S21: data cleaning, specifically, performing missing value processing, outlier detection processing and redundant data elimination operations on the education assessment optimization data set in sequence to obtain cleaned and optimized data;

[0019] Step S22: data fusion, specifically, integrating the cleansed and optimized data, managing the data through primary key matching and timestamp alignment, and processing repeated feature data through manual feature screening to obtain fused and optimized data;

[0020] Step S23: data standardization, specifically, performing data classification standardization on the fusion optimization data, including numerical data standardization, categorical data encoding standardization and text data vector standardization, to obtain normalized comprehensive data, wherein the normalized comprehensive data specifically includes normalized mapping numerical data, numerical encoding normalized data and text vector normalized data;

[0021] The numerical data standardization specifically refers to performing maximum and minimum normalization operations on the numerical data in the fusion optimization data to obtain normalized mapping numerical data;

[0022] The classification data coding standardization specifically refers to performing a unique coding operation on the classification data in the fusion optimization data to obtain numerically coded normalized data;

[0023] The text data vector normalization specifically refers to performing BERT vector conversion on the main text or keyword text data in the fusion optimization data to obtain text vector normalized data;

[0024] Step S24: data enhancement, specifically performing sampling enhancement, data expansion and feature interaction operations on the normalized comprehensive data to obtain an enhanced feature data set;

[0025] The sampling enhancement specifically refers to oversampling the minority category data in the normalized comprehensive data, generating minority category samples by using the SMOTE method, and reducing data redundancy by undersampling the majority category data;

[0026] The data expansion specifically refers to injecting random noise into the learning behavior data and generating new behavior data samples;

[0027] The feature interaction specifically refers to generating combined feature data by combining features, and performing feature statistics to obtain an enhanced feature data set;

[0028] Step S25: data feature optimization, specifically, performing correlation analysis and principal component analysis on the enhanced feature data set, and performing feature dimension reduction to obtain a feature structure optimized data set;

[0029] Step S26: data comprehensive optimization, specifically, performing data comprehensive optimization through the data cleaning, data fusion, data standardization, data enhancement and data feature optimization to obtain an education assessment optimized data set;

[0030] The educational assessment optimization data set specifically includes learning behavior assessment optimization data and educational characteristic assessment optimization data;

[0031] The learning behavior evaluation optimization data specifically includes optimized learning behavior data and text record data;

[0032] The educational characteristic assessment optimization data specifically includes optimized emotional and psychological data.

[0033] Further, in step S3, the learning behavior assessment is used to assess and analyze the students' learning habits and behaviors. Specifically, based on the educational assessment optimization data set, a method combining a causal-inspired enhanced graph neural long-short term memory integrated model is used to conduct learning behavior assessment to obtain learning behavior performance assessment data, which specifically includes the following steps:

[0034] Step S31: constructing a causal heuristic algorithm, specifically using the causal heuristic algorithm to construct a causal graph, and by constructing the causal graph, mining the causal relationship of variables to obtain causal relationship weight data;

[0035] The causal heuristic algorithm specifically refers to a partial correlation algorithm;

[0036] Step S32: constructing a graph neural network, specifically, combining the causal relationship weight data to construct a behavior graph, and extracting student behavior pattern features by constructing a graph neural network to obtain student behavior pattern feature data;

[0037] The behavior graph is constructed by taking the learning behavior variables as nodes and taking the causal relationship weight data as edges;

[0038] Step S33: constructing an improved long short-term memory network, specifically, constructing a capsule network for improving spatial feature modeling to obtain spatial feature output, and jointly modeling the output of the capsule network for improving spatial feature modeling and the temporal feature output of the long short-term memory network to optimize the spatiotemporal dimension of the temporal feature, thereby obtaining comprehensive data for learning behavior evaluation;

[0039] The capsule network for improving spatial feature modeling is constructed by introducing an improved dynamic routing mechanism and using a weighted dynamic routing strategy to optimize the feature selection strategy. The calculation formula is:

[0040] ;

[0041] In the formula, C ij is the dynamic routing weight, which is used to represent the relative importance of student behavior features, i is the input capsule vector index, j is the output capsule vector index, the input capsule vector is used to represent low-level features, the output capsule vector is used to represent high-level features, exp(·) is the natural base function, a ij is the initial weight, K is the normalized traversal total value, k is the normalized traversal index, a ik is the initial weight corresponding to the kth traversal index, u j is the jth output capsule vector, I is the total number of input capsule vectors, and v i is the i-th input capsule vector;

[0042] The long short-term memory network specifically refers to a standard long short-term memory neural network;

[0043] The calculation formula of the joint modeling is:

[0044] ;

[0045] In the formula, ht is the output hidden state of the joint learning behavior evaluation, t is the time index, LSTM(·) is the standard long short-term memory neural network function, V t is the spatial feature output, y t is the prediction result of joint modeling, Softmax(·) is the classifier function, and W o is the output layer weight, b o is the output layer bias term;

[0046] Step S34: training the learning behavior evaluation model, specifically, training the learning behavior evaluation model by constructing the causal heuristic algorithm, constructing the graph neural network, and constructing the improved long short-term memory network to obtain the learning behavior evaluation model Model BP ;

[0047] Step S35: Learning behavior evaluation, specifically using the learning behavior evaluation model Model BP , conducting a learning behavior assessment based on the educational assessment optimization data set to obtain learning behavior performance assessment data;

[0048] The learning behavior performance evaluation data specifically includes quantitative prediction data of classroom attention concentration, classroom discussion participation prediction data, classroom discipline performance prediction data, learning behavior change trend data and comprehensive learning behavior evaluation clustering data.

[0049] Further, in step S4, the educational feature evaluation is used to perform educational feature evaluation and analysis in combination with the learning state information, specifically, based on the learning behavior performance evaluation data and the educational evaluation optimization data set, a graph neural network method combining knowledge transfer and knowledge graph is used to perform educational feature evaluation to obtain educational feature multidimensional evaluation data, specifically including the following steps:

[0050] Step S41: constructing an education feature knowledge graph, specifically by constructing an education feature node set and an education feature edge set to obtain education feature knowledge graph data;

[0051] The educational feature node set specifically includes student nodes, learning behavior nodes, course knowledge mastery nodes, emotional psychology nodes, and student virtue nodes;

[0052] The educational feature edge set specifically refers to the relationship between nodes;

[0053] Step S42: constructing a graph neural network, specifically, constructing a standard graph neural network according to the operation of step S32, and obtaining educational feature node embedding data for representing the relationship characteristics between learning behavior and educational features;

[0054] Step S43: embedding the educational feature knowledge graph, specifically constructing a knowledge graph embedding function, and embedding the educational feature knowledge graph of the graph neural network to obtain an optimized knowledge graph embedded in the graph neural network;

[0055] Step S44: Knowledge transfer prediction, specifically setting the educational feature assessment as the target task, setting the learning behavior performance assessment as the source task, and performing the learning behavior assessment model BP Parameter migration and feature migration, and through knowledge distillation, obtain the knowledge migration prediction model;

[0056] The knowledge distillation is used to use the output of the source task model to guide the learning of the target task model and construct a loss function. The calculation formula is:

[0057] ;

[0058] Where, L KD is the knowledge transfer prediction loss function, which is used to represent the loss function of the education feature evaluation model. is the knowledge transfer loss weight, L CE is the cross entropy loss function of the learning behavior evaluation model, KL(·) is the KL divergence calculation function, P source is the predicted probability output of the source task, which is used to represent the predicted probability output of the learning behavior evaluation model. target is the predicted probability output of the target task, which is used to represent the predicted probability output of the education feature evaluation model;

[0059] Step S45: training the education feature evaluation model, specifically, training the education feature evaluation model by constructing the education feature knowledge graph, constructing the graph neural network, embedding the education feature knowledge graph and predicting the knowledge transfer, to obtain the education feature evaluation model Model EP ;

[0060] Step S46: Educational characteristic evaluation, specifically by using the educational characteristic evaluation model Model EP , performing an educational characteristic evaluation based on the learning behavior performance evaluation data and the educational evaluation optimization data set to obtain educational characteristic multidimensional evaluation data;

[0061] The multi-dimensional evaluation data of educational characteristics specifically include learning behavior dimension data, knowledge mastery dimension data, emotional psychology dimension data, time series trend data, learning virtue assessment data and composite evaluation quantitative data;

[0062] The learning behavior dimension data specifically refers to the attention score, participation score and discipline score of the students’ learning behavior;

[0063] The knowledge mastery dimension data specifically refers to the knowledge mastery degree score of the students in the corresponding subjects;

[0064] The emotional psychological dimension data specifically refers to the students' positivity scores and anxiety scores;

[0065] The time series trend data specifically refers to the time dimension scoring benchmark data, which is used to represent the trend change of the multidimensional evaluation data of educational characteristics over time;

[0066] The learning virtue evaluation data specifically refers to the virtue scoring data of students’ learning behavior habits;

[0067] The composite assessment quantitative data specifically refers to the comprehensive weighted scoring data of learning behavior, knowledge mastery, emotional psychology and learning virtues.

[0068] Furthermore, in step S5, the educational assessment assistance is used to assist in comprehensive educational assessment based on learning behavior and learning status, specifically, to conduct a comprehensive student education assessment in combination with the learning behavior performance assessment data and the educational characteristic multidimensional assessment data, to obtain educational learning behavior assessment results and educational learning characteristic assessment results, and to track and optimize the student education situation based on the educational learning behavior assessment results and the educational learning characteristic assessment results, to obtain updated educational assessment data.

[0069] The present invention provides an artificial intelligence-based education assessment auxiliary system, comprising a data management module, a behavior assessment module, a feature assessment module and an education assessment auxiliary module;

[0070] The data management module is used for data collection and data optimization, and obtains an education assessment optimization data set through data collection and data optimization, and sends the education assessment optimization data set to the behavior assessment module;

[0071] The behavior evaluation module is used for learning behavior evaluation, obtains learning behavior performance evaluation data through learning behavior evaluation, and sends the learning behavior performance evaluation data to the feature evaluation module;

[0072] The feature evaluation module is used for educational feature evaluation, obtains multi-dimensional evaluation data of educational features through educational feature evaluation, and sends the multi-dimensional evaluation data of educational features to the educational evaluation auxiliary module;

[0073] The educational assessment assistance module is used for educational assessment assistance, and through the educational assessment assistance, updated educational assessment data is obtained.

[0074] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0075] (1) In view of the fact that existing educational assessment auxiliary methods often have the problem of single data and single assessment dimension in existing educational assessment intelligent solutions, which leads to the technical problem that the assessment results can neither properly reflect the actual educational status of students nor be effectively intelligent, this solution creatively adopts a two-way correlation analysis method that combines learning behavior and educational characteristics, conducts preliminary analysis through learning behavior assessment, and refines educational characteristics in multiple dimensions, thereby improving the analysis depth and specificity of educational assessment, and at the same time provides an effective way for intelligent methods to assist educational assessment;

[0076] (2) In the existing learning behavior assessment methods, learning behavior often involves complex and deep high- and low-dimensional time-space characteristics. Traditional methods have limited ability to analyze multimodal basic data, which makes it difficult to effectively present the changes in learning behavior in time trends, and it is also impossible to intuitively reflect the results of learning behavior assessment, resulting in low readability, low comprehensibility and low usability. This solution creatively adopts a method of combining causal heuristic enhancement with a graph neural long- and short-term memory integration model to conduct learning behavior assessment. The causal relationship between input data is enhanced through a causal heuristic algorithm, and the time-space characteristics are optimized by constructing an improved capsule network. This improves the prediction accuracy and efficiency of learning behavior assessment as a basic task, improves the effectiveness of the method, and provides a strong model and data foundation for subsequent steps.

[0077] (3) In view of the technical problem that in the existing educational characteristic evaluation methods, there are complex data involving multiple aspects and angles of educational characteristics, and the existing technology cannot effectively decompose and conduct comprehensive evaluation, this solution creatively adopts a graph neural network method that combines knowledge transfer and knowledge graph. By constructing an educational characteristic knowledge graph and embedding it into the graph neural network, it can effectively supplement the educational characteristic analysis information of other dimensions required for learning behavior evaluation. Through transfer learning, it can not only effectively inherit the data results of learning behavior evaluation, but also improve the overall operation efficiency of the model, providing a good operation process for the comprehensive and complex task of assisting educational evaluation. At the same time, the intuitive quantitative data and clear classification data result output also effectively improve the reference and readability of the evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 A flowchart of an artificial intelligence-based educational assessment auxiliary method provided by the present invention;

[0079] Figure 2 A schematic diagram of an artificial intelligence-based education assessment auxiliary system provided by the present invention;

[0080] Figure 3 A schematic diagram of the process for data optimization in step S2;

[0081] Figure 4 This is a flow chart of the learning behavior assessment in step S3;

[0082] Figure 5 Schematic diagram of the process of educational characteristic assessment in step S4.

[0083] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0084] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0085] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0086] Example 1, see Figure 1 The present invention provides an artificial intelligence-based educational assessment auxiliary method, which comprises the following steps:

[0087] Step S1: data collection;

[0088] Step S2: data optimization;

[0089] Step S3: learning behavior assessment;

[0090] Step S4: Educational characteristics assessment;

[0091] Step S5: Educational assessment assistance.

[0092] Example 2, see Figure 1 and Figure 2 , in step S1, the data collection is used to collect the original data set required for educational assessment, specifically, to obtain the original data set for educational assessment through learning scenario behavior data collection;

[0093] The learning scenario behavior data collection specifically includes classroom data collection, extracurricular activity data collection and homework data collection;

[0094] The original educational assessment data set specifically includes learning behavior data, emotional psychology data, and text record data;

[0095] The learning behavior data is specifically collected from the education management system data records, including classroom attention concentration assessment data, classroom discussion participation data and classroom discipline data;

[0096] The classroom attention concentration evaluation data specifically includes the student's classroom attention duration and manual attention score;

[0097] The class discussion participation data specifically includes the number of class discussions, the frequency of class discussions and the duration of class speeches;

[0098] The classroom discipline data specifically includes the number of classroom violations and the content of classroom violations;

[0099] The emotional and psychological data are specifically collected through psychological measurement questionnaires, including students' mental health data and students' emotional state record data;

[0100] The student mental health data specifically includes the student GHQ mental health assessment score data and mental health classification label data; the mental health classification label data specifically includes good, need attention and need special attention;

[0101] The student emotional state record data specifically refers to student emotional state type data and student emotional state change trend data;

[0102] The text record data is specifically collected from homework content and classroom speech records, including student composition homework record data and student classroom speech record data;

[0103] The student composition homework record data specifically includes student composition text data and student composition keyword data;

[0104] The student classroom speech record data specifically includes the student classroom speech text data and the student speech keyword data.

[0105] Example 3, see Figure 1 , Figure 2 and Figure 3 This embodiment is based on the above embodiment. In step S2, the data optimization is used to optimize the original data, specifically to enhance and optimize the original education assessment data set to obtain an education assessment optimized data set, and specifically includes the following steps:

[0106] Step S21: data cleaning, specifically, performing missing value processing, outlier detection processing and redundant data elimination operations on the education assessment optimization data set in sequence to obtain cleaned and optimized data;

[0107] Step S22: data fusion, specifically, integrating the cleansed and optimized data, managing the data through primary key matching and timestamp alignment, and processing repeated feature data through manual feature screening to obtain fused and optimized data;

[0108] Step S23: data standardization, specifically, performing data classification standardization on the fusion optimization data, including numerical data standardization, categorical data encoding standardization and text data vector standardization, to obtain normalized comprehensive data, wherein the normalized comprehensive data specifically includes normalized mapping numerical data, numerical encoding normalized data and text vector normalized data;

[0109] The numerical data standardization specifically refers to performing maximum and minimum normalization operations on the numerical data in the fusion optimization data to obtain normalized mapping numerical data;

[0110] The classification data coding standardization specifically refers to performing a unique coding operation on the classification data in the fusion optimization data to obtain numerically coded normalized data;

[0111] The text data vector normalization specifically refers to performing BERT vector conversion on the main text or keyword text data in the fusion optimization data to obtain text vector normalized data;

[0112] Step S24: data enhancement, specifically performing sampling enhancement, data expansion and feature interaction operations on the normalized comprehensive data to obtain an enhanced feature data set;

[0113] The sampling enhancement specifically refers to oversampling the minority category data in the normalized comprehensive data, generating minority category samples by using the SMOTE method, and reducing data redundancy by undersampling the majority category data;

[0114] The data expansion specifically refers to injecting random noise into the learning behavior data and generating new behavior data samples;

[0115] The feature interaction specifically refers to generating combined feature data by combining features, and performing feature statistics to obtain an enhanced feature data set;

[0116] Step S25: data feature optimization, specifically, performing correlation analysis and principal component analysis on the enhanced feature data set, and performing feature dimension reduction to obtain a feature structure optimized data set;

[0117] Step S26: data comprehensive optimization, specifically, performing data comprehensive optimization through the data cleaning, data fusion, data standardization, data enhancement and data feature optimization to obtain an education assessment optimized data set;

[0118] The educational assessment optimization data set specifically includes learning behavior assessment optimization data and educational characteristic assessment optimization data;

[0119] The learning behavior evaluation optimization data specifically includes optimized learning behavior data and text record data;

[0120] The educational characteristic assessment optimization data specifically includes optimized emotional and psychological data.

[0121] Example 4, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S3, the learning behavior assessment is used to assess and analyze the learning habits and behaviors of students. Specifically, based on the educational assessment optimization data set, a method combining a causal heuristic enhanced graph neural long short-term memory integrated model is used to conduct learning behavior assessment to obtain learning behavior performance assessment data. Specifically, the following steps are included:

[0122] Step S31: constructing a causal heuristic algorithm, specifically using the causal heuristic algorithm to construct a causal graph, and by constructing the causal graph, mining the causal relationship of variables to obtain causal relationship weight data;

[0123] The calculation formula for performing variable causal relationship mining is:

[0124] ;

[0125] In the formula, is a normalized measure used to measure the causal effect of the first variable X on the second variable Y, where the first variable X is specifically used to represent the attention concentration data, and the second variable Y is specifically used to represent the classroom discussion participation data, and Var(·) is the residual variance calculation function;

[0126] The causal heuristic algorithm specifically refers to a partial correlation algorithm;

[0127] Step S32: constructing a graph neural network, specifically, combining the causal relationship weight data to construct a behavior graph, and extracting student behavior pattern features by constructing a graph neural network to obtain student behavior pattern feature data;

[0128] The behavior graph is constructed by taking the learning behavior variables as nodes and taking the causal relationship weight data as edges;

[0129] The calculation formula of the graph neural network is:

[0130] ;

[0131] In the formula, H l+1 is the node feature output of the l+1th graph convolutional layer, which is used to represent the characteristic data output of the student behavior pattern, ReLU(·) is the nonlinear activation function, D is the degree matrix, A is the adjacency matrix, which is used to represent the causal weight matrix, W is the graph neural network weight, and l is the graph convolutional layer index;

[0132] Step S33: constructing an improved long short-term memory network, specifically, constructing a capsule network for improving spatial feature modeling to obtain spatial feature output, and jointly modeling the output of the capsule network for improving spatial feature modeling and the temporal feature output of the long short-term memory network to optimize the spatiotemporal dimension of the temporal feature, thereby obtaining comprehensive data for learning behavior evaluation;

[0133] The capsule network for improving spatial feature modeling is constructed by introducing an improved dynamic routing mechanism and using a weighted dynamic routing strategy to optimize the feature selection strategy. The calculation formula is:

[0134] ;

[0135] In the formula, C ij is the dynamic routing weight, which is used to represent the relative importance of student behavior features, i is the input capsule vector index, j is the output capsule vector index, the input capsule vector is used to represent low-level features, the output capsule vector is used to represent high-level features, exp(·) is the natural base function, a ij is the initial weight, K is the normalized traversal total value, k is the normalized traversal index, a ik is the initial weight corresponding to the kth traversal index, u j is the jth output capsule vector, I is the total number of input capsule vectors, and v i is the i-th input capsule vector;

[0136] The long short-term memory network specifically refers to a standard long short-term memory neural network;

[0137] The calculation formula of the joint modeling is:

[0138] ;

[0139] In the formula, h t is the output hidden state of the joint learning behavior evaluation, t is the time index, LSTM(·) is the standard long short-term memory neural network function, V t is the spatial feature output, y tis the prediction result of joint modeling, Softmax(·) is the classifier function, and W o is the output layer weight, b o is the output layer bias term;

[0140] Step S34: training the learning behavior evaluation model, specifically, training the learning behavior evaluation model by constructing the causal heuristic algorithm, constructing the graph neural network, and constructing the improved long short-term memory network to obtain the learning behavior evaluation model Model BP ;

[0141] Step S35: Learning behavior evaluation, specifically using the learning behavior evaluation model Model BP , conducting a learning behavior assessment based on the educational assessment optimization data set to obtain learning behavior performance assessment data;

[0142] The learning behavior performance evaluation data specifically includes quantitative prediction data of classroom attention concentration, classroom discussion participation prediction data, classroom discipline performance prediction data, learning behavior change trend data and comprehensive learning behavior evaluation clustering data.

[0143] By performing the above operations, in view of the fact that in the existing learning behavior evaluation methods, learning behaviors often involve complex and deep high and low-dimensional time and space characteristics, and traditional methods have limited analysis capabilities for multimodal basic data, which makes it difficult to effectively present the changes in learning behavior in time trends, and it is also impossible to intuitively reflect the results of learning behavior evaluation, resulting in low readability, low comprehensibility and low usability. Technical problems, this solution creatively adopts a method combining causal inspiration enhanced graph neural long short-term memory integrated model to conduct learning behavior evaluation, enhances the causality between input data through causal inspiration algorithm, and optimizes time and space characteristics by constructing an improved capsule network, thereby improving the prediction accuracy and prediction efficiency of learning behavior evaluation as a basic task, improving the effectiveness of the method and providing a strong model and data foundation for subsequent steps.

[0144] Example 5, see Figure 1 , Figure 2 and Figure 5 This embodiment is based on the above embodiment. In step S4, the educational feature evaluation is used to evaluate and analyze the educational features in combination with the learning state information. Specifically, based on the learning behavior performance evaluation data and the educational evaluation optimization data set, a graph neural network method combining knowledge transfer and knowledge graph is used to evaluate the educational features to obtain multi-dimensional evaluation data of educational features. Specifically, the following steps are included:

[0145] Step S41: constructing an education feature knowledge graph, specifically by constructing an education feature node set and an education feature edge set to obtain education feature knowledge graph data;

[0146] The educational feature node set specifically includes student nodes, learning behavior nodes, course knowledge mastery nodes, emotional psychology nodes, and student virtue nodes;

[0147] The educational feature edge set specifically refers to the relationship between nodes;

[0148] Step S42: constructing a graph neural network, specifically, constructing a standard graph neural network according to the operation of step S32, and obtaining educational feature node embedding data for representing the relationship characteristics between learning behavior and educational features;

[0149] Step S43: embedding the educational feature knowledge graph, specifically constructing a knowledge graph embedding function, and embedding the educational feature knowledge graph of the graph neural network to obtain an optimized knowledge graph embedded in the graph neural network;

[0150] The calculation formula of the knowledge graph embedding function is:

[0151] ;

[0152] In the formula, Score(·) is the score of the knowledge graph embedding function, which is used to optimize the knowledge graph embedding. is the head node, used to represent the student learning node, R is the relationship node, used to represent the knowledge mastery and emotional psychological state, T is the tail node, used to represent the learning morality, ||·||2 is the L2 norm calculation operator;

[0153] Step S44: Knowledge transfer prediction, specifically setting the educational feature assessment as the target task, setting the learning behavior performance assessment as the source task, and performing the learning behavior assessment model BP Parameter migration and feature migration, and through knowledge distillation, obtain the knowledge migration prediction model;

[0154] The knowledge distillation is used to use the output of the source task model to guide the learning of the target task model and construct a loss function. The calculation formula is:

[0155] ;

[0156] Where, L KD is the knowledge transfer prediction loss function, which is used to represent the loss function of the education feature evaluation model. is the knowledge transfer loss weight, L CE is the cross entropy loss function of the learning behavior evaluation model, KL(·) is the KL divergence calculation function, P source is the predicted probability output of the source task, which is used to represent the predicted probability output of the learning behavior evaluation model. targetis the predicted probability output of the target task, which is used to represent the predicted probability output of the education feature evaluation model;

[0157] Step S45: training the education feature evaluation model, specifically, training the education feature evaluation model by constructing the education feature knowledge graph, constructing the graph neural network, embedding the education feature knowledge graph and predicting the knowledge transfer, to obtain the education feature evaluation model Model EP ;

[0158] Step S46: Educational characteristic evaluation, specifically by using the educational characteristic evaluation model Model EP , performing an educational characteristic evaluation based on the learning behavior performance evaluation data and the educational evaluation optimization data set to obtain educational characteristic multidimensional evaluation data;

[0159] The multi-dimensional evaluation data of educational characteristics specifically include learning behavior dimension data, knowledge mastery dimension data, emotional psychology dimension data, time series trend data, learning virtue assessment data and composite evaluation quantitative data;

[0160] The learning behavior dimension data specifically refers to the attention score, participation score and discipline score of the students’ learning behavior;

[0161] The knowledge mastery dimension data specifically refers to the knowledge mastery degree score of the students in the corresponding subjects;

[0162] The emotional psychological dimension data specifically refers to the students' positivity scores and anxiety scores;

[0163] The time series trend data specifically refers to the time dimension scoring benchmark data, which is used to represent the trend change of the multidimensional evaluation data of educational characteristics over time;

[0164] The learning virtue evaluation data specifically refers to the virtue scoring data of students’ learning behavior habits;

[0165] The composite assessment quantitative data specifically refers to the comprehensive weighted scoring data of learning behavior, knowledge mastery, emotional psychology and learning virtues.

[0166] By performing the above operations, in order to address the technical problem that in the existing educational feature evaluation methods, there are complex data involving multiple aspects and angles of educational features, and the existing technology cannot effectively decompose and conduct comprehensive evaluation, this solution creatively adopts a graph neural network method that combines knowledge transfer and knowledge graph. By constructing an educational feature knowledge graph and embedding it into the graph neural network, it can effectively supplement the educational feature analysis information of other dimensions required for learning behavior evaluation. Through transfer learning, it can not only effectively inherit the data results of learning behavior evaluation, but also improve the overall operation efficiency of the model, providing a good operation process for the comprehensive and complex task of assisting educational evaluation. At the same time, the intuitive quantitative data and clear classification data result output also effectively improve the reference and readability of the evaluation results.

[0167] Example 6, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S5, the educational assessment assistance is used to assist in comprehensive educational assessment based on learning behavior and learning status. Specifically, it combines the learning behavior performance assessment data and the educational characteristic multidimensional assessment data to conduct a comprehensive assessment of student education, obtain educational learning behavior assessment results and educational learning characteristic assessment results, and track and optimize the student education situation based on the educational learning behavior assessment results and educational learning characteristic assessment results to obtain updated educational assessment data.

[0168] Embodiment 7, see Figure 1 and Figure 2 , this embodiment is based on the above embodiment, and the present invention provides an artificial intelligence-based education assessment auxiliary system, including a data management module, a behavior assessment module, a feature assessment module and an education assessment auxiliary module;

[0169] The data management module is used for data collection and data optimization, and obtains an education assessment optimization data set through data collection and data optimization, and sends the education assessment optimization data set to the behavior assessment module;

[0170] The behavior evaluation module is used for learning behavior evaluation, obtains learning behavior performance evaluation data through learning behavior evaluation, and sends the learning behavior performance evaluation data to the feature evaluation module;

[0171] The feature evaluation module is used for educational feature evaluation, obtains multi-dimensional evaluation data of educational features through educational feature evaluation, and sends the multi-dimensional evaluation data of educational features to the educational evaluation auxiliary module;

[0172] The educational assessment assistance module is used for educational assessment assistance, and through the educational assessment assistance, updated educational assessment data is obtained.

[0173] By executing the above operations, in view of the problem that the existing educational assessment auxiliary methods often have the problem of single data and single assessment dimension in the existing educational assessment intelligent solutions, which leads to the technical problem that the assessment results can neither properly reflect the actual educational status of students nor be effectively intelligent, this solution creatively adopts a two-way correlation analysis method combining learning behavior and educational characteristics, conducts preliminary analysis through learning behavior assessment, and refines educational characteristics through multi-dimensional means, which enhances the analysis depth and specificity of educational assessment, and at the same time provides an effective way for intelligent methods to assist educational assessment.

[0174] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0175] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0176] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. An artificial intelligence-based educational assessment auxiliary method, characterized in that: The method comprises the following steps: Step S1: Data collection, obtaining an original data set of educational assessment by collecting learning scenario behavior data; the learning scenario behavior data collection specifically includes classroom data collection, extracurricular activity data collection and homework data collection; the original data set of educational assessment specifically includes learning behavior data, emotional psychology data and text record data; Step S2: data optimization, enhancing and optimizing the original education assessment data set to obtain an education assessment optimized data set; Step S3: learning behavior assessment, used to assess and analyze students' learning habits and behaviors, specifically based on the educational assessment optimization data set, using a method of a graph neural long-term and short-term memory integrated model combined with causal inspiration enhancement to conduct learning behavior assessment and obtain learning behavior performance assessment data, specifically including the following steps: step S31: constructing a causal inspiration algorithm; step S32: constructing a graph neural network; step S33: constructing an improved long-term and short-term memory network, specifically by constructing a capsule network for improved spatial feature modeling to obtain spatial feature output, and by jointly modeling the output of the capsule network for improved spatial feature modeling and the temporal feature output of the long-term and short-term memory network, the spatiotemporal dimension of the temporal feature is optimized to obtain comprehensive learning behavior assessment data; the construction of the capsule network for improved spatial feature modeling specifically refers to introducing an improved dynamic routing mechanism and using a weighted dynamic routing strategy to optimize the feature selection strategy; step S34: learning behavior assessment model training; step S35: learning behavior assessment; Step S4: educational feature evaluation, which is used to evaluate and analyze educational features in combination with learning status information, specifically, based on the learning behavior performance evaluation data and the educational evaluation optimization data set, a graph neural network method combining knowledge transfer and knowledge graph is used to evaluate educational features, and obtain multi-dimensional evaluation data of educational features; Step S5: Educational assessment assistance, combining the learning behavior performance assessment data and the educational characteristic multi-dimensional assessment data, to conduct a comprehensive assessment of student education, and to track and optimize the student education situation to obtain updated educational assessment data.

2. The artificial intelligence-based educational assessment auxiliary method according to claim 1, characterized in that: In step S1, the learning behavior data is specifically collected from the education management system data records, including classroom attention concentration evaluation data, classroom discussion participation data and classroom discipline data; The emotional and psychological data are specifically collected through psychological measurement questionnaires, including students' mental health data and students' emotional state record data; The text record data is specifically collected from homework content and classroom speech records, including student composition homework record data and student classroom speech record data.

3. The educational assessment auxiliary method based on artificial intelligence according to claim 2, characterized in that: In step S31, the causal heuristic algorithm is constructed, specifically, a causal graph is constructed using the causal heuristic algorithm, and variable causal relationship mining is performed by constructing the causal graph to obtain causal relationship weight data; In step S32, the construction of the graph neural network is specifically to construct a behavior graph in combination with the causal relationship weight data, and extract student behavior pattern features by constructing the graph neural network to obtain student behavior pattern feature data; In step S33, the calculation formula for constructing the capsule network for improving spatial feature modeling is: ； In the formula, C ij is the dynamic routing weight, which is used to represent the relative importance of student behavior features, i is the input capsule vector index, j is the output capsule vector index, the input capsule vector is used to represent low-level features, the output capsule vector is used to represent high-level features, exp(·) is the natural base function, a ij is the initial weight, K is the normalized traversal total value, k is the normalized traversal index, a ik is the initial weight corresponding to the kth traversal index, u j is the jth output capsule vector, I is the total number of input capsule vectors, and v i is the i-th input capsule vector; The long short-term memory network specifically refers to a standard long short-term memory neural network; The calculation formula of the joint modeling is: ； In the formula, h t is the output hidden state of the joint learning behavior evaluation, t is the time index, LSTM(·) is the standard long short-term memory neural network function, V t is the spatial feature output, y t is the prediction result of joint modeling, Softmax(·) is the classifier function, and W o is the output layer weight, b o is the output layer bias term; In step S34, the learning behavior evaluation model training is specifically carried out by constructing the causal heuristic algorithm, the graph neural network and the improved long short-term memory network to obtain the learning behavior evaluation model Model BP ; In step S35, the learning behavior assessment is specifically performed using the learning behavior assessment model Model BP , conducting a learning behavior assessment based on the educational assessment optimization data set to obtain learning behavior performance assessment data; The learning behavior performance evaluation data specifically includes quantitative prediction data of classroom attention concentration, classroom discussion participation prediction data, classroom discipline performance prediction data, learning behavior change trend data and comprehensive learning behavior evaluation clustering data.

4. The artificial intelligence-based educational assessment auxiliary method according to claim 3, characterized in that: In step S4, the education feature assessment specifically includes the following steps: Step S41: constructing an education feature knowledge graph, specifically by constructing an education feature node set and an education feature edge set to obtain education feature knowledge graph data; The educational feature node set specifically includes student nodes, learning behavior nodes, course knowledge mastery nodes, emotional psychology nodes, and student virtue nodes; The educational feature edge set specifically refers to the relationship between nodes; Step S42: constructing a graph neural network, specifically, constructing a standard graph neural network according to the operation of step S32, and obtaining educational feature node embedding data for representing the relationship characteristics between learning behavior and educational features; Step S43: embedding the educational feature knowledge graph, specifically constructing a knowledge graph embedding function, and embedding the educational feature knowledge graph of the graph neural network to obtain an optimized knowledge graph embedded in the graph neural network; Step S44: Knowledge transfer prediction, specifically setting the educational feature assessment as the target task, setting the learning behavior performance assessment as the source task, and performing the learning behavior assessment model BP Parameter migration and feature migration, and through knowledge distillation, obtain the knowledge migration prediction model; Step S45: training the education feature evaluation model, specifically, training the education feature evaluation model by constructing the education feature knowledge graph, constructing the graph neural network, embedding the education feature knowledge graph and predicting the knowledge transfer, to obtain the education feature evaluation model Model EP ; Step S46: Educational characteristic evaluation, specifically by using the educational characteristic evaluation model Model EP , based on the learning behavior performance evaluation data and the education evaluation optimization data set, an education characteristic evaluation is performed to obtain multi-dimensional education characteristic evaluation data.

5. The artificial intelligence-based educational assessment auxiliary method according to claim 4, characterized in that: In step S44, the knowledge distillation is used to use the output of the source task model to guide the learning of the target task model and construct a loss function, and the calculation formula is: ； Where, L KD is the knowledge transfer prediction loss function, which is used to represent the loss function of the education feature evaluation model. is the knowledge transfer loss weight, L CE is the cross entropy loss function of the learning behavior evaluation model, KL(·) is the KL divergence calculation function, P source is the predicted probability output of the source task, which is used to represent the predicted probability output of the learning behavior evaluation model. target It is the predicted probability output of the target task, which is used to represent the predicted probability output of the education feature evaluation model.

6. The artificial intelligence-based educational assessment auxiliary method according to claim 5, characterized in that: In step S46, the educational characteristic multi-dimensional evaluation data specifically includes learning behavior dimension data, knowledge mastery dimension data, emotional psychology dimension data, time series trend data, learning virtue evaluation data and composite evaluation quantitative data; The learning behavior dimension data specifically refers to the attention score, participation score and discipline score of the students’ learning behavior; The knowledge mastery dimension data specifically refers to the knowledge mastery degree score of the students in the corresponding subjects; The emotional psychological dimension data specifically refers to the students' positivity scores and anxiety scores; The time series trend data specifically refers to the time dimension scoring benchmark data, which is used to represent the trend change of the multidimensional evaluation data of educational characteristics over time; The learning virtue evaluation data specifically refers to the virtue scoring data of students’ learning behavior habits; The composite assessment quantitative data specifically refers to the comprehensive weighted scoring data of learning behavior, knowledge mastery, emotional psychology and learning virtues.

7. The artificial intelligence-based educational assessment auxiliary method according to claim 6, characterized in that: In step S5, the educational assessment assistance is used to assist in comprehensive educational assessment based on learning behavior and learning status, specifically, to conduct a comprehensive student education assessment in combination with the learning behavior performance assessment data and the educational characteristic multidimensional assessment data, to obtain educational learning behavior assessment results and educational learning characteristic assessment results, and to track and optimize the student education situation based on the educational learning behavior assessment results and educational learning characteristic assessment results, to obtain updated educational assessment data.

8. An artificial intelligence-based education assessment auxiliary system, used to implement an artificial intelligence-based education assessment auxiliary method as described in any one of claims 1 to 7, characterized in that: It includes data management module, behavior assessment module, feature assessment module and education assessment auxiliary module.

9. The artificial intelligence-based education assessment auxiliary system according to claim 8, characterized in that: The data management module is used for data collection and data optimization, and obtains an education assessment optimization data set through data collection and data optimization, and sends the education assessment optimization data set to the behavior assessment module; The behavior evaluation module is used for learning behavior evaluation, obtains learning behavior performance evaluation data through learning behavior evaluation, and sends the learning behavior performance evaluation data to the feature evaluation module; The feature evaluation module is used for educational feature evaluation, obtains multi-dimensional evaluation data of educational features through educational feature evaluation, and sends the multi-dimensional evaluation data of educational features to the educational evaluation auxiliary module; The educational assessment assistance module is used for educational assessment assistance, and through the educational assessment assistance, updated educational assessment data is obtained.

Images