A basic education enhancement system based on deep learning

Abstract

The present invention relates to a system for enhancing basic education based on deep learning, belonging to the field of educational informatization. The system includes: an external information perception module, a perception memory module, a working memory module, a longterm memory module, an activation condition module, an acquisition condition module, a consolidation condition module, a transfer condition module, a quantitative data processing module, and a results visualization module. The quantitative data processing module includes: an SPSS analysis module, a DINA cognitive diagnosis module, a hierarchical analysis module, and a multifaceted Rasch module. The present invention is highly operable and contributes to improving teachers' teaching abilities and professional levels, scientifically promoting the enhancement of basic education quality.

Description

l A BASIC EDUCATION ENHANCEMENT SYSTEM BASED ON DEEP LEARNING TECHNICAL FIELD The present invention belongs to the eld of educational informatization and relates to a basic education enhancement system based on deep learning. BACKGROUND TECHNOLOGY The current knowledge processing mechanism theory in China has the following deciencies: (1) Most of the proposed teaching conditions for the chemical knowledge processing stages are theoretical explanations, without corresponding teaching experiments to verify whether these conditions can truly achieve efcient teaching outcomes. Most studies are merely theoretical, lacking sufcient connection with actual teaching practices. The proposed measures have weak operability and practicality in teaching, making it difcult for them to have a positive guiding role in actual teaching. (2) There is a lack of in-depth exploration of the knowledge processing mechanism that affects efcient teaching and the impact of these factors on the teaching conditions across the four stages of chemical knowledge processing. Currently, domestic research on efcient teaching has a narrow scope regarding inuencing factors, with no specic teaching experiments conducted to explore these factors, leading to insufcient theoretical research and lack of evaluation results. This makes it difcult to apply the proposed teaching conditions in practice. (3) Research on the knowledge transfer stage in chemical knowledge processing is still relatively undeveloped. In the study of teaching conditions across the four stages of knowledge processing, there is relatively little research related to the knowledge transfer stage. As a result, the proposed teaching conditions are insufcient, leaving a research gap in this stage and causing the loss of instructional guidance. Therefore, there is an urgent need in China to develop a more detailed and practically guiding system for improving basic education teaching efciency based on deep learning. SUMMARY OF THE INVENTION In view of the above, the purpose of this invention is to provide a basic education enhancement system based on deep learning, addressing the low operability of conditions in the four stages of knowledge processing, which makes the proposed teaching conditions difcult to apply in practice. To achieve the above purpose, as shown in Figures 1- 2, the invention provides the following technical solution: A basic education enhancement system based on deep learning, comprising: an external information perception module, a perception memory module, a working memory module, a long-term memory module, an activation condition module, an acquisition condition module, a consolidation condition module, a transfer condition module, a quantitative data processing module, and a results visualization module; The quantitative data processing module includes: an SPSS analysis module, a DINA cognitive diagnosis module, a hierarchical analysis module, and a multifaceted Rasch module; The external information perception module inputs information into the perception memory module upon reaching the threshold attention level. If the information in the perception memory module resonates with the information in the long-term memory module after being activated by the activation condition module and undergoing pattern recognition and matching, the information in the perception memory module is amplied and directly enters the long-term memory module, externally manifesting as automated efcient learning; If the information in the perception memory module is too strong or does not match the information in the long-term memory module, it is input into the working memory module. The information in the working memory module undergoes deep processing, consolidation, and transfer as required by the acquisition condition module and consolidation and transfer condition module of deep learning, matching and combining with the information in the long- term memory module, and then entering the long-term memory module. Externally, this manifests as the student mastering knowledge through deep learning; When knowledge is retrieved, the information in the long-term memory module is activated and transferred to the working memory module by the activation condition module, externally manifesting as the student having a good recollection of the knowledge; The activation conditions in the activation condition module include: activation through the integration of new and old knowledge, activation through situational cues, and activation through prediction and open analysis; The conditions in the acquisition condition module that strengthen information include: a classroom atmosphere of high challenge and low stress, deep processing of information, and the proceduralization of knowledge based on declarative knowledge. The deep processing of information includes: content processing, structural processing, and connection and positioning processing; content processing involves rening concepts using declarative and procedural denitions; structural processing involves triple representation through information representation, organization, and "macro-micro symbols"; Knowledge declarative includes: pattern recognition, concept acquisition, and information visualization; Knowledge proceduralization includes: knowledge decomposition, knowledge connection and positioning, and knowledge practice feedback; The conditions in the consolidation condition module for consolidating information include: deliberate practice and variation practice, spaced practice and interleaved practice, collaborative situational task practice, massed practice and distributed practice, physical practice, and mental practice; The conditions in the transfer condition module for transferring information include: model-based processing, which extracts substantive information from knowledge, identies causal logic between concepts, generalizes concepts, and forms intellectual models in space, time, causality, logic, and quantication; systematic problem solving, which involves correctly describing the problem, analyzing the problem, constructing solutions, evaluating solutions, and developing solutions based on the acquired information, and reasonably interpreting data from different sources to acquire systematic problem-solving abilities; metacognitive feedback practice, which involves supervising one's performance and thinking to promote self-behavior and thought, including feedback on subject knowledge, task knowledge, and strategy knowledge; The SPSS analysis module quanties the conditions for activating information in the activation condition module, representing the quantitative results through the results visualization module, and judging whether the information has been activated based on the quantitative results represented in the results visualization module; The DINA cognitive diagnosis module quanties the conditions for acquiring information in the acquisition condition module, representing the quantitative results through the results visualization module, and judging whether the information has been acquired based on the quantitative results represented in the results visualization module; The hierarchical analysis module quanties the conditions for consolidating information in the consolidation condition module, representing the quantitative results through the results visualization module, and judging whether the information has been consolidated based on the quantitative results represented in the results visualization module; The multifaceted Rasch module quanties the conditions for transferring information in the transfer condition module, representing the quantitative results through the visualization module, and judging whether the information has been transferred based on the quantitative results represented in the visualization module. Furthermore, the SPSS analysis module quanties the conditions for activating information through SPSS software; The conditions for activating information are quantied through SPSS software, specically by: The SPSS data preparation stage, where the functions provided are used to prepare data les, including dening the structure of SPSS data, entering and modifying it in the data editor window; The SPSS data processing and organization stage, where the data in the data editor window is preprocessed; The SPSS data analysis stage, where the correct analysis methods are chosen for the data and regression analysis modeling is conducted, with SPSS software autonomously modeling and automatically calculating the results; The interpretation stage of analysis results, where the meanings of the calculated results are claried, and reasonable interpretations are made in conjunction with actual background knowledge. Furthermore, the DINA cognitive diagnosis module quanties the conditions for acquiring information through the DINA model; the DINA model is expressed as follows P(x = 1|a-) = (1 s)"""g;'"°' ( K m,- = ÍÍ- k: 1 s, = P(x = OI) = 1) 9j=P(XU=1|'7iJ=0) (2) Where xi,- refers to the score of student i on item j, ai refers to the mastery level of student i on the knowledge points, (zik refers to the mastery level of student i on knowledge point k, and qjk refers to the examination of knowledge point k by item j; ni; equals l or 0; ni; = 1 indicates that the examinee i has mastered all the attributes tested by item j , while ni; = 0 indicates that the examinee i has not mastered at least one attribute tested by item j ; Sj is referred to as the item slip parameter, indicating the probability that the examinee incorrectly answers the item despite having the required attributes to solve item j due to a mistake; g j is referred to as the item guessing parameter, indicating the probability that the examinee correctly answers the item despite not fully mastering the necessary attributes to solve item j due to a lucky guess; to ensure model identiability, for each item, P(Xij = l l ni; = l) Z P(Xij = l l ni; = 0); the accuracy of the model will be inuenced by S)- and gj. Constructing an adjacency matrix A and a reachability matrix R, obtaining a full event matrix Q from the reachability matrix R, deriving a reduced event matrix Qr from the full event matrix Q, and obtaining the Ideal Mastery Pattern (IMP) from the reduced event matrix Qr. A diagnostic test framework is formed based on the results of matrix calculations. Constructing the adjacency matrix A is based on whether there is a direct logical relationship between pairs of attributes; if a direct relationship exists, it is represented by 1; otherwise, it is represented by 0. In the hierarchical relationship of cognitive attributes in the "Periodic Law," Al is directly related to A2 and A3, A2 is directly related to A5, A3 is directly related to A5, A4 is directly related to A5, and A5 is directly related to A6. Therefore, the adjacency matrix based on the hierarchical relationship of attributes in this study is shown in Table 1. Table l - Adjacency Matrix A Attributes .--.| A} A3 A4 AS AE: AI II] | l II] II] II] A} II] II] II] II] | II] A] II] II] II] II] | II] 3.4 II] II] II] II] | II] 5.5 II] II] II] II] II] | .äñ II] II] II] II] II] II] _ The reachability matrix R is constructed based on whether there is a direct, indirect, or self-relationship between pairs of attributes; if such a relationship exists, it is represented by 1; otherwise, it is represented by 0. In the hierarchical relationship of cognitive attributes in the Periodic Law," for example, A4 has a self-relationship with A4, a direct relationship with A5, and an indirect relationship with A6, so all are represented by "1," while the rest are represented by "0." Accordingly, the reachability matrix based on the hierarchical relationship of attributes in the "Periodic Law can be constructed, as shown in Table 2. Table 2 - Reachability Matrix R AttribUtes .4l .4: .43 .44 .45 .45 .-'4| | | | II] | | HQ II] | II] II] | | Hú II] II] | II] | | .it-l II] II] II] | | | .-'-.."'I II] II] II] II] | | Haß II] II] II] II] II] | _ Based on the R matrix, the expansion algorithm is applied to obtain the full event matrix Q, which reects the relationship between attributes and possible item type sets. The full event matrix Q contains a total of 26 = 64 possible item sets. Items that do not conform to the logical relationships between attributes are removed, and the column that examined O attributes is deleted, resulting in the reduced event matrix Qr, which forms the framework for the cognitive diagnostic test, as shown in Table 3. Table 3 - Reduced Event Matrix Qr Possible Items Attributes | 3 3 4 6 '.l 3 9 |IIII | l Al | l | IIII | l l | | | l A3 IIII l Ü IIII | l IIII | | IIII l A] II] II] | IIII | l IIII | IIII | l A] II] II] II] | | l l IIII | | l A5 II] II] II] II] | l II] II] II] II] II] A6 II] II] II] II] II] l II] II] II] II] II] The reduced event matrix Qr is transposed to obtain the transposed matrix Qr'. This matrix theoretically reects the attribute mastery patterns of the examinees that conform to the hierarchical relationships of attributes, known as the Ideal Mastery Pattern, also referred to as the knowledge state. On the basis of the transposed matrix Qr', a row of all zeros is added, resulting in the Ideal Mastery Pattern as shown in Table 4. Table 4 - Ideal Mastery Pattern Attribut Ideal - Examinee A | 9.3 A3 A4 9.5 9.5 | II] II] II] II] II] II] É | II] II] II] II] II] 3 | | II] II] II] II] 4 | II] | II] II] II] 5 II] II] II] | II] II] ô | | | | | II] Î" | | | | | | 3 | II] II] | II] II] 9 | | | II] II] II] III] | | II] | II] II] | | | II] | | II] II] IE | | | | II] II] The students response data is encoded in a data table with correct responses coded as 1 and incorrect responses as 0, and saved as a CSV le. The encoded student response data and the corresponding Q matrix are uploaded to the cognitive diagnostic analysis platform, where a suitable DINA model is selected to obtain the students attribute mastery patterns and attribute mastery probabilities. Furthermore, the hierarchical analysis module quanties the conditions required for information consolidation through a hierarchical analysis model. The hierarchical structure model, as shown in Figure 1, includes the highest layer, middle layer, and lowest layer. For two adjacent layers, the higher layer is referred to as the goal layer, and the lower layer as the factor layer. The factors within the same layer belong to or inuence the factors of the upper layer, while also controlling or being inuenced by the factors of the lower layer. The topmost layer is the goal layer, typically consisting of only one factor, representing the objective of solving the problem. The middle layer consists of one or several criteria that must be followed when choosing measures or plans to achieve the overall goal, also known as the strategy layer, constraint layer, or criterion layer. When there are more than 9 criteria, they are further decomposed into sub-criteria layers. The lowest layer is the plan or object layer. Furthermore, as shown in Figure 2, the hierarchical analysis model is used to quantify the conditions required for information consolidation, specically including: Constructing judgment matrices A, starting from the second layer of the hierarchical structure model. For the factors within the same layer that belong to or inuence each factor of the upper layer, judgment matrices A are constructed until the lowest layer. When determining the weights between the factors of each layer, the consistency matrix method is used, which involves: Comparing factors pairwise rather than all together to reduce the difculty of comparing factors with different natures and to improve accuracy. Using a relative scale when comparing factors pairwise to minimize the complexity and to ensure accuracy. The judgment matrix A represents the relative importance comparison of all factors within the current layer relative to a specic factor in the upper layer. The elements a.,- of the judgment matrix A are given using Satty's l-9 scale method, where: A strength of 1 indicates that two factors are of equal importance. A strength of 3 indicates that one factor is slightly more important than the other. A strength of 5 indicates that one factor is signicantly more important than the other. A strength of 7 indicates that one factor is more important than the other. A strength of 9 indicates that one factor is extremely more important than the other. Strengths of 2, 4, 6, and 8 represent intermediate levels of importance. The reciprocal indicates that if factor i is judged to be b times more important than factor j, then factor j is judged to be l / b times as important as factor i. The judgment matrix A is expressed as: szXn... ... ... ... (4) The elements in A satisfy the conditions ai,- a.,- = 1 / a, and a..- = 1. This process continues until a complete judgment matrix A is established. Once the judgment matrix is constructed, hierarchical single sorting and its consistency test are performed. Hierarchical single sorting involves pairwise comparison of a specic element in the upper layer with all elements in the current layer, followed by hierarchical sorting to determine the order of importance. The sorting calculation is based on the judgment matrix A, ensuring that the judgment matrix Ameets the conditions of eigenvalues and eigenvectors, specically: AW = ÂmaxW ( 5 ) The largest eigenvalue of matrix A is denoted as Àmax, and the normalized eigenvector corresponding to Àmax is W. The component W.- of W represents the weight, which corresponds to the single sorting of its respective element. The weight of each factor a.,- in the judgment matrix A relative to the goal layer is calculated by determining the weight vector W and the largest eigenvalue Àmax. The steps for calculating the weight vector W and the largest eigenvalue Xmax are as follows: First, compute the product of the elements in each row, then take the n-th root of the product to obtain an n-dimensional vector, which can be expressed by the formula: H": n jH3=1a,-j 1,]=1. 2. 3......n (6) Normalize W.- so that the sum of all elements in the vector equals 1. The formula for normalization is: w. Wi = n' Zi: .l. W] (7) After normalizing wi, you obtain the ranking weight vector, denoted as W. The elements in W represent the ranking weights of factors within the same layer relative to a specic factor in the upper layer. Therefore, W = (W1, W2, Wn)T is the weight vector that needs to be determined and also the result of the hierarchical single sorting of the judgment matrix. The largest eigenvalue of the judgment matrix A is given by the formula: _1 (AW I . ÄmaxízrzlTl (8) To solve for the largest eigenvalue and the consistency index CI value: Let the n-order judgment matrix be B, and then the largest eigenvalue Àmax is calculated as follows: BW = ÀW (9) Where W is the eigenvector of B. In the Analytic Hierarchy Process (AHP), the consistency index CI is used to test the consistency of the judgments, and it is expressed as: 0.1. : Em_" n = 1 (10) CI. = 0 indicates that the judgment matrix is completely consistent, and the larger the C.I. value, the more severe the inconsistency of the judgment matrix. The consistency ratio CR for the hierarchical single sorting is calculated using the CI and RI values, and it is expressed as: G.I. C.R. = R.I. ( ll ) The random consistency index R.I values for matrices of orders l-l3 are as follows: When the matrix order is l, the RI value is 0 ; When the matrix order is 2, the RI value is 0 ; When the matrix order is 3, the RI value is 0.58; When the matrix order is 4, the RI value is 0.90; When the matrix order is 5, the R.I value is 1.12; When the matrix order is 6, the RI value is 1.24; When the matrix order is 7, the RI value is 1.32; When the matrix order is 8, the R.I value is 1.41; When the matrix order is 9, the R.I value is 1.45; When the matrix order is 10, the R.I value is 1.49; When the matrix order is 11, the R.I value is 1.51; When the matrix order is 12, the R.I value is 1.54; When the matrix order is 13, the R.I value is 1.56; When C.R. < 0.1, it indicates that the consistency of the judgment matrix A is considered within the acceptable range, and the eigenvector of A is used to calculate the weight vector. When C.R. 2 0.1, the judgment matrix A should be revised. Hierarchical overall sorting and its consistency test: Hierarchical overall sorting involves calculating the relative importance weights of all factors at a certain level with respect to the highest level. This process is carried out from the highest level to the lowest level: Let A be the highest level, including m factors, with the overall sorting weight coefcients being al \ az. ag, ..., a. LetBbe the middle level, including n factors, with the single sorting weight coefcients being b}......b." Then z),-:zglajbi, represents the overall sorting of level B. Let the consistency indices for the hierarchical sorting of factors B1, B2, ..., Bn in level B with respect to factors A,- in level A, where j = l, 2, ..., m, be CI j , and the random consistency indices be RIj. Then, the consistency ratio for the overall sorting is: l6 CR 0,1011 + a2012+m + aCIm E (WI. CI a1R11+ G2RI2 + ' " + amRIm E (z,-RI, RI (12) When C.R. < 0.1, the overall sorting is considered to have passed the consistency test; otherwise, the elements of the judgment matrix need to be adjusted, and the nal decision is made based on the overall sorting of the lowest level, i.e., the decision layer. Furthermore, the multifaceted Rasch module quanties the conditions required for information transfer using the Rasch model. The Rasch model is expressed as: P (Xmi=1|6 6,-)=exp (6,,1-6i) / [1+exp (B-&)] (13) P (Xmi =l|6 (î,-) refers to the probability that an individual with ability 9 correctly answers an item x = 1 with difculty (S.-. The Rasch model quanties the conditions required for information transfer, specically including: The Rasch model evaluates item difculty levels and student ability levels based on test result data, placing student ability levels and item difculty levels on the same interval scale for comparison. Reliability is mainly assessed based on item and examinee reliability, error, and discrimination indices. Validity is mainly tested based on data-model t, including unidimensionality, item-examinee correspondence, data-model t indices MNSQ, ZSTD, and point-measure correlation PTMEA indices for analysis and evaluation. Data results are analyzed using Winsteps 4.4.0 software to conduct overall quality testing. The overall quality testing chart reects the overall t as well as reliability and separation issues. Raw scores are converted to l7 logit values, and a Wright map is used to place student abilities and items on the same scale, allowing for an intuitive and simple comparison of item difculty with student abilities, as well as a comparative analysis of different item difculties. In the Wright map, the leftmost numbers represent the logit interval scale values used to compare examinee ability levels and item difculty. The logit values increase from bottom to top, indicating higher examinee ability levels and increased item difculty. Ideally, each item in the test corresponds to an examinee with a matching ability level, and there should be more items set in areas where examinees are relatively concentrated. The vertical line in the middle is the logit scale, representing the measured symbol ability structure. M, S, and T represent the mean level, one logit, and two logits, respectively. The right side of the scale shows the distribution of test items. When testing item t, Winsteps provides two forms of chi-square t indices: Outt MNSQ represents the mean square residual, and Int MNSQ represents the weighted mean square residual. The range of Outt MNSQ and Int MNSQ values is from 0 to positive innity, with an ideal value of 1, indicating that the actual data ts the model well. A value greater than 1 (Undert) indicates that the variance of the empirical data is higher than expected; a value less than 1 (Overt) indicates that the variance of the empirical data is lower than expected. Furthermore, items with problematic t are handled by deleting or revising the problematic items, modifying the scoring criteria, and analyzing the factors inuencing item issues. The analysis of factors inuencing item issues includes examining aspects such as discrimination and unidimensionality. The benecial effects of the present invention include: In terms of data processing, the present invention exibly uses domestic and foreign computer processing software for quantitative analysis, such as constructing models using software like Facet and Conquest, using cognitive diagnostic matrices like Q matrix and R matrix, conducting evaluations on cognitive diagnostic analysis platforms (exCDMs), and using Winsteps software for analyzing scale reliability, t, and other aspects. The present invention evaluates efcient teaching conditions based on knowledge processing mechanisms by employing a combination of analytic hierarchy process (AHP), multifaceted Rasch model analysis, DINA cognitive diagnosis model, and SPSS analysis. The combined use of multiple analytical methods provides high reliability and operability for the experimental quantication in the present invention. The present invention utilizes multiple evaluation tools and methods to reasonably evaluate teaching conditions from multiple dimensions, emphasizing the importance of teachers in organizing and using knowledge. It advocates for the careful renement of concepts through declarative and procedural denitions, with a thorough understanding of declarative knowledge as a prerequisite for learning procedural knowledge. Procedural knowledge is the basis for knowledge transfer, so during teaching, it is recommended to attempt using procedural knowledge in the learning of declarative knowledge. Before learning new knowledge, it is important to have a solid grasp of prior knowledge and to engage in both massed practice and distributed practice. Teachers should consciously use intellectual models to address different problem situations and apply knowledge transfer through analogy and generalization, thereby creating an efcient teaching environment. The present invention is highly operable and benecial for improving teachers teaching abilities and professional levels. Other advantages, objectives, and features of the present invention will be explained to some extent in the following description, and to some extent, they will be apparent to those skilled in the art from the study of the following or may be learned from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained by the explanations provided below. DESCRIPTION OF THE DRAWINGS To make the objectives, technical solutions, and advantages of the present invention clearer, the following preferred detailed descriptions are provided in conjunction with the accompanying drawings, wherein: Figure 1 is a hierarchical evaluation model framework diagram; Figure 2 is a data processing owchart within the hierarchical evaluation model; Figure 3 is a bar chart showing the overall distribution of students mastery of six cognitive attributes; Figure 4 is a radar chart of the probability of attribute mastery for certain examinees; Figure 4(a) is the radar chart of attribute mastery probability for examinee 44; Figure 4(b) is the radar chart of attribute mastery probability for examinee 100; Figure 4(c) is the radar chart of attribute mastery probability for examinee 128; Figure 4(d) is the radar chart of attribute mastery probability for examinee 336; Figure 4(e) is the radar chart of attribute mastery probability for examinee 346; Figure 4(f) is the radar chart of attribute mastery probability for examinee 419; Figure 5 is a hierarchical structure diagram of evaluation indicators for chemistry evidence reasoning ability among high school students in ethnic regions; Figure 6 is an overall quality inspection chart obtained from the overall quality testing of data for each examinee using corresponding software; Figure 7 is a Wright map; Figure 8 is a unidimensionality plot; Figure 9 is a system diagram of the present invention. o DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following describes the embodiments of the present invention through specic examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specication. The present invention can also be implemented or applied through other different specic embodiments, and various modications or changes can be made to the details in this specication without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only for explaining the basic concept of the present invention in a schematic manner, and where there is no conict, the following embodiments and the features in the embodiments can be combined with each other. The accompanying drawings are merely for illustrative purposes and represent schematic diagrams, not actual product diagrams, and should not be understood as limiting the present invention. In order to better explain the embodiments of the present invention, certain components in the drawings may be omitted, enlarged, or reduced, and do not represent the actual size of the products. It is understandable that certain well-known structures and their descriptions may be omitted from the drawings for those skilled in the art. In the drawings of the embodiments of the present invention, identical or similar reference numbers correspond to identical or similar components. In the description of the present invention, it should be understood that if terms such as upper, lower, left, right," front, back, etc., are used to indicate orientation or positional relationships, they are based on the orientation or positional relationships shown in the drawings, and are merely for the convenience of describing the present invention and simplifying the description. They are not intended to indicate or imply that the referred devices or elements must have a specic orientation or be constructed and operated in a specic orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be understood as limiting the present invention. Those skilled in the art can understand the specic meanings of the above terms based on specic situations. Please refer to Figures 3 to 8 for a system for enhancing basic education based on deep learning. In this system, the quantitative data module quanties the conditions in each condition module. The quantitative results displayed in the results visualization module are used to determine whether the information has been activated, acquired, consolidated, and transferred. |. Case Study on the Quantitative Analysis of Conditions for Activating Information Using SPSS Analysis Module The recognition of teaching situations created based on the knowledge processing mechanism and the achievement of the three- dimensional objectives after class were assessed for the ninth-grade students of a private middle schoolSchool Nin Dazu District, Chongqing. Two classes with equivalent levels were selected as the experimental class and the control class, with a total of 60 students. SPSS 21.0 software was used to analyze the collected data, including student performance analysis and student questionnaire analysis. A. Student Performance Analysis 1. Analysis of Pre-test Scores of Experimental and Control Classes Before conducting various statistical analysis methods, it is necessary to perform a normality test on the data. There are two methods for normality testing in SPSS: the Kolmogorov-Smimov test (K-S test) and the Shapiro-Wilk test (S-W test). The pre-test chemistry scores of the two classes from the previous semester were imported into SPSS 21.0, and an independent sample t-test was used to test the differences. The results of the normality distribution test of the t-test are shown in Table 5. Table 5 - Normality Test of Scores of Experimental and Control Classes (Pre-test) Kolmogorov-Smimova Shapiro-Wilk Perfor Statistic df Sig. Statistics df Sig. mance S Control 0.118 30 0.200"< 0.954 30 0.210 Group Experi 0.088 30 0.200"< 0.977 30 0.736 mental Group If the data sample size is between 3 and 50, we tend to rely on the results of the Shapiro-Wilk (S-W) test. Since the number of students in both the experimental class and the control class is 30, which falls into the small sample size category, we use the S-W test results: for the experimental class, Sig = 0.736 > 0.05; for the control class, Sig = 0.210 > 0.05. It can be concluded that the scores of both the experimental class and the control class follow a normal distribution. Therefore, we proceed with the mean difference analysis of the scores for the two classes. The test results are shown in Table 6. Table 6 - Group Statistics of Scores for the Experimental and Control Classes (Pre-test) Class N Mean Standard Standard Deviation Error of Mean Perfor Control 30 39.550 13.8691 2.5321 mance Group Experime 30 41.500 10.9308 1.9957 ntal Group As shown in Table 6, the average score for the control class is 39.55, and the average score for the experimental class is 41.50. It can be seen that the mean difference in scores between the two classes is not signicant, indicating that the overall level of the two classes is quite similar. To test whether this difference is statistically signicant, we further analyze the output results of the independent sample t-test, as shown in Table 7. Table 7 - Independent Sample Test of Scores for the Experimental and Control Classes (Pre-test) Levenes Test for Equality of t-test for Equality of Means Variances F Sig t df Sig Mean Standard 95% Condence (2- Differ Error Interval of the taile ence Differenc e leference d) Lower Upper Bound Bound I'erfor Assumed 1.0 0.3 - 58 0.54 - 3.2240 - 4.5036 ance Equal 10 19 0.60 8 1.950 8.4036 Variance 5 Assumed - 54.9 0.54 - 3.2240 - 4.5111 Unequal 0.60 9 8 1.950 8.411] Variance 5 In conducting a two-sample t-test, an F-test is used to determine whether the variances of the two samples are equal, known as the homogeneity of variance. Therefore, an F-test (also known as the homogeneity of variance test) is performed rst. According to the t-test results in Table 7, the signicance value (Sig) next to F is 0.319 > 0.05, indicating that there is no signicant difference in the variances of the two classes. Therefore, the results of the t-test should be interpreted based on the rst row of data, assuming equal variances. If the signicance value Sig of the F-test were less than 0.05, it would indicate no signicant difference between the two groups. Consequently, it can be concluded that there is no signicant difference in the scores of the two classes, further conrming that the experimental class and control class are of similar levels and can be considered homogeneous for subsequent experimental studies. 2. Analysis of Post-test Scores for the Experimental and Control Classes After the teaching was completed, both the control class and the experimental class took the nal test. The scores of the students in the experimental class and the control class were imported into SPSS 21.0 software. First, the normality test was conducted on the test scores of students in both classes, and the test results are shown in Table 8. Table 8 - Normality Test of Scores for the Experimental and Control Classes (Post-test) .. Kolmogorov-Smimova Shapiro-Wilk orm Control 0.091 30 0200* 0.953 30 0.203 ance Group Experi 0.085 30 0200* 0.965 30 0.406 mental Group Since the sample size for both classes is 30, which is considered a small sample, the Shapiro-Wilk test results are preferred. For the control class, Sig. = 0.203 > 0.05; for the experimental class, Sig. = 0.406 > 0.05. These tests indicate that the scores of both the experimental class and the control class follow a normal distribution, allowing us to proceed with the mean difference analysis between the two classes. The analysis results are shown in Table 9. Table 9 - Group Statistics of Scores for the Experimental and Control Classes (Post-test) Class N Mean Standard Standard Deviation Error of Mean Perfor Control 30 42.983 15.0645 2.7504 mance Group Experime 30 50.017 11.5292 2.1049 ntal Group As shown in Table 9, the mean score of students in the experimental class is 50.017, while the mean score of students in the control class is 42.983. The mean score of the experimental class is signicantly higher than that of the control class. Next, an independent sample t-test was conducted on the scores of the experimental and control classes, and the results are shown in Table 10. Table 10 - Independent Sample Test of Scores for the Experimental and Control Classes (Post-test) Levenes Test for Equality of Variances t-test for Equality of Means F Sig t df Sig Mean Standard 95% Condence (2- Differ Error Interval of the taile ence Differenc e Difference d) Lower Upper Bound Bound I'erfor Assumed 2.03 0.15 -.2.03 58 0.04 - 3.4634 - -0.1005 mance Equal 2 9 1 7 7.033 13.966 Variances 2 Assumed -.2.03 54.2 0.04 - 3.4634 - -0.0904 Unequal 1 9 7 7.033 13.976 As shown in Table 10, the F value from the homogeneity of variances 5 test is 2.032, with Sig = 0.159 > 0.05, which satises the assumption of equal variances. The corresponding P value is 0.047 < 0.05, indicating a signicant difference in test scores between the experimental class and the control class at the 0.05 signicance level. This further proves that the teaching situations created based on the knowledge processing mechanism have a signicant impact on students' learning. B. Student Questionnaire Analysis A questionnaire was prepared to understand students' perceptions of chemistry classroom teaching based on the knowledge processing mechanism and to assess the achievement of knowledge, emotional, and ability objectives. Thirty questionnaires were collected, and the reliability and validity of the questionnaire were analyzed using SPSS 21.0 software. 1. Reliability Table 11 - Reliability Statistics Table As shown in Table 11, the Cronbachs Alpha reliability coefcient for the student questionnaire is 0.925, indicating that the reliability of the survey results is very high. 2. Validity Table 12 - KMO and Bartlett's Test Kaiser-Meyer-Olkin Measure 0 0.823 Sampling Adequacy Bartlett's Test of Sphericity Approximate Chi- 270.242 Square 0.000 Scholar Kaiser suggested that when conducting factor analysis, the KMO value should be at least above 0.6. As shown in Table 12, the KMO value of this questionnaire is 0.823 > 0.8, indicating that the questionnaire suitable for factor analysis. Factor analysis is the most commonly used method to test the construct validity of a questionnaire. Construct validity simply refers to whether there are overlapping or cross-loading factors (which can be understood as items) among the questionnaire items. Factor analysis extracts the main common factors. The results of the factor analysis for this student questionnaire are shown in Table 13. Table 13 - Total Variance Explained Com Initial Eigenvalues Extraction Sums of Rotation Sums of Squared Squared Loadings Loadings onent Total Percenta Cumul Total Percenta Cumul Total Percenta Cumu ge of a tive ge of a tive ge of l ative Variance % Variance % Variance % 5.280 65.994 65.994 5.280 65.994 65.994 3.368 42.104 42.10 4 1.114 14.303 80.297 1.144 14.303 80.297 3.055 38.194 80.29 7 0.083 1.036 100.00 0 Extraction Method: Principal Component Analysis According to the total variance explained in Table 13, it can be observed that there are 2 components with initial eigenvalues greater than 1, and the cumulative total variance is 80.297% > 80%, indicating a 5 relatively high contribution rate. Therefore, extracting 2 factors in the factor analysis is appropriate. Subsequently, the orthogonal rotation method with Kaiser normalization is used to obtain the rotated component matrix, as shown in Table 14. Table 14 - Rotated Component Matrix Based on the rotated component matrix in Table 14, it can be seen that the two components are quite reasonable. ||. Case Study on the Quantitative Analysis of Conditions for Obtaining Information Using the DINA Cognitive Diagnosis Module A total of 436 rst-year high school students from a high school in Changsha, Hunan Province, were selected as the study subjects. The nal test was conducted using a paper- and-pencil test method. The data collected was analyzed using the G-DINA model analysis toolCognitive Diagnosis Analysis Platform (exCDMs), primarily focusing on identifying the knowledge states of the examinees and the overall attribute mastery ratios. A. Identication of Examinee Knowledge States Based on the G-DINA Model Using the Cognitive Diagnosis Analysis Platform (exCDMs), the G- DINA model was selected, and the classication results of the examinees were obtained using the Expected A Posteriori (EAP) estimation method. classication statistics of the knowledge states (KS) of the examinees shown in Table 15. Table 15 - Classication Statistics of Examinee Knowledge States ID / Number Attribute Mastery Number of Classie Classication Pattern Individuals / People Rate / % ___ Based on Table 15, it can be observed that out of 430 examinees, 391 examinees' actual response patterns could be classied into 12 knowledge states, achieving a classication rate of 90.9%, which meets Tasuoka's standard of 90%. This indicates that the G-DINA model can be successfully applied to the cognitive diagnostic assessment of "Periodic Law content for rst-year high school students. To more clearly understand the overall distribution of students' mastery of the six cognitive attributes, a bar chart was created as shown in Figure 3. From Figure 3, it can be seen that the response patterns of most examinees are classied into KS7 (111000), KS8 (111100), KS9 (111110), and KS10 (111 1 11), indicating that most students have at least mastered the three attributes of atomic structure (A1), periodic table of elements (A2), and periodic law of elements (A3). Among them, the largest number of students, nearly one-sixth, are classied into KS7, mastering the three attributes of atomic structure (A1), periodic table of elements (A2), and periodic law of elements (A3). The number of students classied into KS9 (111110) ranks second, with approximately 13.7% of students mastering ve attributes: atomic structure (Al), periodic table of elements (A2), periodic law of elements (A3), properties and uses of substances based on representative substances and classication views (A4), and the application of periodic table and periodic law of elements (A5). More than 10% of students are classied into KS8 (1 1 1 100), indicating that about one- tenth of students have mastered the four attributes of atomic structure (Al), periodic table of elements (A2), periodic law of elements (A3), and properties and uses of substances based on representative substances and classication views (A4). A. Overall Attribute Mastery Ratios Based on the G-DINA Model Using the Cognitive Diagnosis Analysis Platform (exCDMs), the G- DINA model was selected, and the posterior distribution estimation method (EAP) was applied to obtain the mastery ratios of the cognitive attributes. The mastery ratios of the six cognitive attributes related to the Periodic Law" content for the 430 examinees are shown in Table 16. Table 16 - Statistical Results of Attribute Mastery Ratios for the Periodic Law Mastery 94.9% 68.2% 73.6% 61.4% 31.9% 19.4% Rate / % Overall, the mastery ratios of students for the six attributes of the "Periodic Law knowledge content, from highest to lowest, are as follows: atomic structure (Al) > periodic law of elements (A3) > periodic table of elements (A2) > properties and uses of substances based on representative substances and classication views (A4) > application of periodic table and periodic law of elements (A5) > properties and uses of substances based on the "position-structure-property" understanding of elements (A6). Notably, less than half of the examinees mastered attributes A5 and A6. In other words, 68.1% of the examinees were unable to use the position of elements in the periodic table and atomic structure to analyze, predict, and compare the properties of elements, and only 19.4% of the examinees could predict the properties and uses of unknown elements. B. Radar Chart of Attribute Mastery Probabilities for Examinees Based on the G-DINA Model In addition to obtaining the overall attribute mastery ratios of the examinees, the Cognitive Diagnosis Platform (exCDMs) can also be used to obtain the mastery probability of each individual for each attribute based on the G-DINA model and generate corresponding radar charts of attribute mastery probabilities. Six examinees with relatively typical attribute mastery patterns were selected. Their attribute mastery patterns are shown in Table 17, and the radar charts of attribute mastery probabilities are shown in Figure 4. Table 17 - Attribute Mastery Patterns for Selected Examinees Examinee Attribute Mastery Examinee Attribute Mastery ID / Number Pattern ID / Number Pattern ID44 110100 ID336 111111 ID100 110000 ID346 111100 ID128 110000 ID419 111110 For examinee ID 44, the mastery probabilities for the six attributes are 1.0000, 0.9895, 0.0141, 1.0000, 0.1446, and 0.0027, respectively. Combining this with the radar chart in Figure 4(a) for examinee ID 44, it can be more intuitively seen that examinee ID 44 has a good grasp of attributes A1 and A4, while the mastery probabilities for attributes A2, A3, A5, and A6 are lower, with the mastery probability for attribute A6 reaching only 0.0027. For examinee ID 100, the mastery probabilities for the six attributes are 1.0000, 0.9988, 0.1894, 0.0293, 0.0067, and 0.0081, respectively. Combining this with the radar chart in Figure 4(b) for examinee ID 100, it can be more intuitively seen that examinee ID 100 has a good grasp of attributes A1 and A2, while the mastery probabilities for attributes A3, A4, A5 , and A6 are lower. For examinee ID 128, the mastery probabilities for the six attributes are 0.9988, 0.5233, 0.1306, 0.0108, 0.4768, and 0.0048, respectively. Combining this with the radar chart in Figure 4(c) for examinee ID 128, it can be more intuitively seen that examinee ID 128 has a good grasp of attribute A1 , while the mastery probabilities for attributes A2, A3, and A5 are lower, with the mastery probability for attribute A6 reaching only 0.0048. For examinee ID 336, the mastery probabilities for the six attributes are 1.0000, 0.9811, 0.9840, 0.9996, 0.9735, and 0.9649, respectively. Combining this with the radar chart in Figure 4(d) for examinee ID 336, it can be more intuitively seen that examinee ID 336 has fully mastered the six attributes of the "Periodic Law," with the mastery probabilities for all six attributes being close to 1. For examinee ID 346, the mastery probabilities for the six attributes are 1.0000, 0.8282, 0.9989, 1.0000, 0.0029, and 0.1597, respectively. Combining this with the radar chart in Figure 4(e) for examinee ID 346, it can be more intuitively seen that examinee ID 346 has a good grasp of attributes A1 , A2, A3, and A4, while the mastery probabilities for attributes A5 and A6 are lower. For examinee ID 419, the mastery probabilities for the six attributes are 1.0000, 0.6043, 0.6055, 1.0000, 0.6043, and 0.0012, respectively. Combining this with the radar chart in Figure 4(f) for examinee ID 419, it can be more intuitively seen that examinee ID 419 has the best grasp of attributes A1 and A4, followed by A2, A3, and A5, with a lower mastery probability for attribute A6. It can be seen that the radar charts of attribute mastery probabilities help teachers more clearly and intuitively understand students' knowledge states, allowing teachers to take targeted remedial teaching measures based on students strengths and weaknesses. |||. Case Study on the Quantitative Analysis of Conditions Required for Information Consolidation Using the Hierarchical Analysis Module Constructing the Hierarchical Structure Model Based on the evaluation indicators for the reasoning ability of high school students in ethnic regions, a hierarchical structure model of the evaluation was constructed in Yaahp, as shown in Figure 5. Constructing the Judgment Matrix Using Saaty's 1-9 scale method as shown in Table 18, this study designed an expert questionnaire to conduct pairwise comparisons of the importance of 3 primary indicators and 6 secondary indicators. Table 18 - Saaty's Fundamental Scale Table Importance Denition Description Intensity 1 Equal Indicates that two factors are equally important whe Importance compared 3 Slightly More Indicates that one factor is slightly more important Important than the other when compared 5 Clearly More Indicates that one factor is clearly more important tha Important the other when compared 7 More Indicates that one factor is more important than the Important other when compared Extremely Indicates that one factor is extremely more importan- Important than the other when compared 2,4,6,8 Intermediate Indicates transitional importance between mai Values Between judgments Judgments Reciprocal If factor i is judged to be b times more important than factor j, then factor j is judged to be 1 / b times more important than factor i According to Saaty's 1-9 scale method shown in Table 18, this study designed an expert questionnaire for pairwise comparisons of the importance of 3 primary indicators and 6 secondary indicators. C. Judgment Matrix Calculation 1. Consistency Test Due to differences in cognitive levels, there are also differences in understanding objective matters, so when experts judge the importance of indicators, inconsistencies or lack of coordination between the importance of elements may occur. If the degree of inconsistency reaches a certain level, it will affect the accuracy of the judgment. Therefore, to improve the accuracy of weight measurement, this study uses Saaty's consistency test Zx n (ÍÍ _ method. The consistency index n _ ] of the judgment matrix, and the average random consistency index RI varies with the matrix order n, and its values are shown in Table 19. The ratio of the consistency index CI to R1 is called the random consistency ratio CR of the judgment matrix. When CR S 0.1, the judgment matrix is considered to have satisfactory consistency. Table 19 - Average Random Consistency Index RI Values Matrix 1 2 3 4 5 7 10 11 12 13 Order n R.I. 0.5 1.1 1.2 1.3 1.4 1.4 1.4 1.5 1.5 1.5 8 2 4 2 1 5 9 1 4 6 The results of 155 valid questionnaires were entered into Yaahp for this study, and 58 of these questionnaires passed the consistency test. Using the Geometric Mean Method (Square Root Method) to Calculate the Relative Weight Values Wi l ( : ai ) m=HW.]19 i=1.2...'...l n n __ " Ì . (l_l . "U) 1=1 j=1 (1) @ First, calculate the product of the elements in each row; @ Then, take the nth root of the product for each row; @ Normalize the resulting vector to obtain the weight vector value for each expert; @ Calculate the total weight value by averaging the weight vector values of all experts to obtain the average weight value for the indicator system, as shown in Table 20: Table 20 - Distribution of Evaluation Indicator Weights Primary Weight Secondary Same-Level Global Indicators Indicators Weight Weight Collect 0.3187 Collect Information 0.4153 0.1323 Chemical . Identify Evidence 0.5 847 0.1863 EVidence Formulate 0.3534 Interpret 0.473 0.1672 Hypotheses Evidence Based on _ Formulate 0.527 0.1863 EVidence Hypotheses Draw 03279 Find Logical 0.4794 0.1572 Conclusions Relationships Through Validate 0.5206 0.1707 Reasoning Conclusions Through Reasoning The evaluation model for the chemistry evidence reasoning ability of high school students in ethnic regions, determined using the Analytic Hierarchy Process (AHP), can be mathematically represented as: Y=0.32A+0.35B+0.33C (2) Where A=0.42a1+0.58a2, B=0.47b1+0.53b2, C=0.48c1+0.52c2 Therefore Y=0.13a1+0.19a2+0.17b1+0.19b2+0.16c1+0.17c2o V. Case Study on the Quantitative Analysis of Conditions Required for Information Transfer Using the Multifaceted Rasch Module The cognitive ability in chemistry modeling of high school seniors at two ordinary high schools, S High School in Xinjiang and M High School in Shandong, was assessed. A total of 170 students, with 85 students from each school, participated in the test. Both schools are ordinary county-level high schools, and the students come from parallel classes. The tool used was the multifaceted Rasch model analysis toolWinsteps 4.4.0 software. After 30 students completed the test, the data were collected and analyzed, focusing on aspects such as the unidimensionality of the scale, t, overall test performance of students, and score variability. A Overall Quality Measurement The data from 170 examinees were subjected to overall quality testing using Winsteps 4.4.0 software, as shown in Figure 6. The results show that the MNSQ values for items and examinees are close to the ideal value of 1, and the ZSTD values are within an acceptable range. The overall reliability of the items (Item reliability = 0.98, > 0.7) and the overall reliability of the examinees (Person reliability = 0.75, > 0.7) are both high. The item separation (Tap separation = 8.02, > 2) indicates that the test items can distinguish between examinees of different levels. The Wright map, as shown in Figure 7, presents the Logit scale values on the far left, which indicate the comparison between the examinee's ability level and the difculty of the items. From top to bottom, the Logit values increase, indicating an increase in both the examinee's ability level and the difculty of the items. The distribution of item difculty is relatively wide and generally even; the distribution of student ability is also ideal, with a certain range, being more concentrated in the middle and less at the ends, basically showing a normal distribution. Items SC41, SC42, and SC43 with Logit values above 2 have no corresponding examinees, which is related to the fact that the examinees come from parallel classes in ordinary high schools in Shandong and Xinjiang, where the overall student level is not high. It can also be seen that the examinees in this study have relatively low abilities in the fourth level of mathematical modeling: comprehensive application level, with almost no students reaching this level. B Fit Table 21 shows that the range of item difculty spans from -2.12 to 3.60, which is slightly reduced compared to the pilot test. This indicates that the measured items are relatively stable in assessing the cognitive ability in modeling of high school seniors. The estimation error of item difculty has signicantly decreased compared to the pilot test. Except for the estimation error of item SC43, which is 0.51, the errors for other items are around 0.1, indicating that the estimated parameters of the measured items are relatively stable. Compared to the pilot test, where the item difculty errors were around 0.5, the errors are now reduced to around 0.1, indicating that the modication of the items has reduced the difference between observed and expected values, thereby improving the accuracy of measurement. It also suggests that the relatively large estimation errors in the pilot test were due to the small sample size. Except for item SC3, where the ZSTD value slightly exceeds the range, the Int and Outt MNSQ and ZSTD values for other items are ideal. The "point- measure correlation coefcients are all between 0 and 1, with no negative values, meeting the requirements, and there are 9 items with "point-measure data above 0.35, accounting for half of the total measured items, indicating that the items have good discriminatory ability and differentiation effects. Table 21 - Fit Table mw'w-W- e A wawa mmmmmmmm mammmmnm mmmmmmmm Mmmmmmmm ammmmmmm ammmmmmm mmmmmmmm mmmmmmmm mmmmmmmm ammmmmmm ammmmmmm mmmmnmmm mmmmmmmm anmmmmmm Emmmmmmm Mmmmmmmm Unidimensionality As shown in Figure 8, the standardized residual contrast loading for most items falls within the range of -0.4 to 0.4. Except for items a (GBl), A (SC42), and B (SC41), which are outside this range, the majority of items fall within the ideal range. Although a few items are outside the range, they are not far from it, with only item SC42 being positioned further from the ideal range, which is considered acceptable. Compared to the pilot test, this test more accurately reects the current level of the examinees cognitive ability in modeling, indicating good unidimensionality. Analysis of Overall Student Test Scores Table 22 - Statistical Results of Overall Student Test Scores All Count Person Item S.E. Item Lable Score Measure Item Measure refers to the estimated difculty level of an item, indicating the specic difculty level of the item (Rasch score). An analysis table of overall student test scores was created by plotting the data of total number of students, student ability, item difculty, error, and item number. Through the overall item difculty analysis of student assessment results from the two schools, the ability levels of students from the two schools across different dimensions were compared. The data analysis results show that for the students from both schools, the measurement errors of item difculty are mostly concentrated within the range of 0.1 to 0.26, indicating small measurement errors in item difculty. By calculating the average item difculty for each dimension, it was found that the average item difculty for physical models is 060, for conceptual models is - 0.50, and for mathematical models is 1.19. The analysis of overall student test scores reveals that the current level of cognitive ability in modeling for high school seniors in the two schools, ranked from high to low, is as follows: physical models, conceptual models, and mathematical models. Analysis of Student Test Score Differences Any measurement (exam) consists of specic items, and differences in cultural background, living environment, and other factors may affect the familiarity and understanding ability of examinees with respect to test items, potentially leading to different results. This can result in measurement outcomes that are favorable to some groups or individuals while being unfavorable to others, leading to the phenomenon of differential item functioning (DIF). The formula for DIF is expressed as follows: p (x|6 g) =p (x|9) (3) where 0 represents the latent trait of the examinees, and g is the group indicator value. For all 0 and g, x is the observed variable. Generally, the absence of DIF can be understood as a form of conditional independence, meaning that the probability of a certain response does not depend on the group membership of the examinees. Table 23 - Statistical Results of Examinee Ability in Different Schools The data from Shandong M High School and Xinjiang S High School were encoded as SD and XJ, respectively. As shown in Table 23, for the 170 students from Shandong and Xinjiang who participated in the overall test, the average ability level is -0.48, indicating a relatively low ability level; the error is 0.96, which is relatively small; the separation index is 1.86, close to 2, indicating that the items can effectively differentiate between high-ability and low-ability examinees; the reliability is 0.78 > 0.7, indicating that the test tool measures the current level of students cognitive ability in modeling well, and the assessment results are relatively ideal. The average ability of students from Shandong M High School is -0.44, with an error of 1.11, and a separation index of 2.07 > 2, indicating that the assessment can effectively differentiate between high-ability and low-ability students in Shandong M High School; the reliability is 0.81 > 0.7, indicating that the assessment results are consistent, reliable, and stable. The ability assessment results for students from Xinjiang S High School show an ability value of -0.52, with an error of 0.80, which is smaller than the assessment error of 1.11 for students from Shandong M High School, indicating that there is a certain difference between observed values and expected values, and the test tool has a higher precision in assessing students from Xinjiang S High School compared to Shandong M High School. The separation index for examinees from Xinjiang S High School is 1.55, indicating a moderate separation; the reliability for students from Xinjiang S High School is 0.71 > 0.7, which is relatively ideal. This indicates that the measurement tool performs well in assessing the cognitive ability in modeling for students from Xinjiang S High School. Table 24 - Statistical Table of Differences in Cognitive Ability Levels for Different Schools The ability value for Shandong M High School is -0.44, while for Xinjiang S High School, it is -0.52, resulting in an ability difference of 0.08. This indicates that the cognitive ability in modeling for students from Shandong M High School is slightly higher than that of students from Xinjiang S High School. A P-value less than 0.05 indicates a signicant difference; in this assessment, the t-value is 0.57, and the Prob. is 0.569, indicating no signicant difference. Conclusion: The cognitive ability in modeling for high school seniors at Shandong M High School is slightly higher than that of high school seniors at Xinjiang S High School, but there is no signicant difference between the two. Table 25 - Analysis Table of Differences in Test Item Difculty for Different Schools item SD XJ XJ DIF S.E. t Prob. DIF DIF DIF Contr Meas Meas S'E' ast ure ure WA11 -1.98 0.30 -2.25 0.30 0.27 0.42 0.65 0.515 2 WA12 -1.36 0.26 -2.08 0.28 0.72 0.38 1.87 0.063 1 WA21 -2.72 0.36 -1.40 0.25 -1.32 0.44 -3.01 0.003 1 WA22 1.22 0.33 1.16 0.32 0.46 0.13 0.895 9 mammmmmnw mmmmnmmaw anmnmmnnm aanammmmw ammnnmnmw anmmmmmmw From the DIF difference analysis between Shandong M High School and Xinjiang S High School, it can be seen that in the items WA21, WA4, GB1, GB21, GB32, SCl, SC41, and SC42, the Prob. value is less than 0.05, indicating a signicant difference. For the rst item of understanding the physical model at M High School, the cognitive ability of high school seniors at M High School is signicantly higher than that of high school seniors at S High School; however, for the comprehensive application level of the physical model, it is signicantly lower than that of S High School students. For the rst item of memory and understanding levels of the conceptual model, M High School students ability is signicantly higher than that of S High School students; for the second item of simple application level, M High School students ability is signicantly lower than that of S High School students. For the memory level of the mathematical model, M High School students ability is signicantly higher than that of S High School students; for the rst two items of comprehensive application level, M High School students' ability is signicantly lower than that of S High School students. Lastly, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modications or equivalent replacements can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, all of which should be covered within the scope of the claims of the present invention. PREFERRED EMBODIMENTS 1. A basic education enhancement system based on deep learning, comprising: an external information perception module, a perception memory module, a working memory module, a long-term memory module, an activation condition module, an acquisition condition module, a consolidation condition module, a transfer condition module, a quantitative data processing module, and a result visualization module; wherein the quantitative data processing module comprises an SPSS analysis module, a DINA cognitive diagnosis module, a hierarchical analysis module, and a multifaceted Rasch module; wherein the external information perception module, upon reaching an attention threshold, inputs information into the perception memory module; if information in the perception memory module is activated by the activation condition module and undergoes pattern recognition and matching, and then resonates with information in the long-term memory module, the information in the perception memory module is amplied and directly enters the long-term memory module, with the external manifestation being the generation of automated and efcient learning; if processing of information in the perception memory module is excessive or the information does not match information in the long-term memory module, the information is input into the working memory module, wherein information in the working memory module, through deep processing, consolidation, and transfer in accordance with the acquisition condition module and the consolidation and transfer condition modules required for deep learning, is matched and combined with information in the long-term memory module, and then enters the long-term memory module, with the external manifestation being that a student has mastered knowledge through deep processing learning; when knowledge is retrieved, information in the long-term memory module is activated by the activation condition module and transferred into the working memory module, with the external manifestation being that a student recalls knowledge well; wherein the conditions for activating information in the activation condition module comprise newold knowledge association integration activation initiation, situational cue activation initiation, and prediction and open-ended analysis activation initiation; wherein the conditions for strengthening information in the acquisition condition module comprise: a high-challenge and low-alert classroom atmosphere, deep processing of information, and proceduralization based on knowledge statement; wherein the deep processing of information comprises: content-based processing, structural processing, and connection positioning processing; wherein the content-based processing comprises rening concepts using declarative denitions and procedural denitions; wherein the structural processing comprises information representation and organization, and macromicro symbol triple representation; wherein the knowledge statement comprises pattern discrimination, concept acquisition, and information diagramming; wherein the knowledge proceduralization comprises overall decomposition of knowledge, knowledge connection positioning, and knowledge practice feedback; wherein the conditions for consolidating information in the consolidation condition module comprise deliberate practice and variation practice, spaced practice and interleaved connection, collaborative situational task practice, massed practice and distributed practice, and physical practice and mental practice; wherein the conditions for transferring information in the transfer condition module comprise model-based processing, namely extracting substantive information of knowledge, identifying causal logic between concepts, generalizing concepts, and thereby forming spatial, temporal, causal, logical, and quantitative intellectual models; systematic problem solving, namely applying acquired information to correctly describe problems, analyze problems, construct solutions, evaluate solutions, and develop solutions, as well as reasonably interpreting data from different sources to obtain the capability of systematic problem solving; and metacognitive feedback practice, namely monitoring ones performance and thinking to promote ones own behavior and thought, including providing feedback on the subjects knowledge, feedback on task knowledge, and feedback on strategy knowledge; wherein the SPSS analysis module performs quantitative processing on the conditions for activating information in the activation condition module, outputs the quantied results through the result visualization module, and determines whether information has been activated according to the quantied results displayed in the result visualization module; wherein the DINA cognitive diagnosis module performs quantitative processing on the conditions for acquiring information in the acquisition condition module, outputs the quantied results through the result visualization module, and determines whether information has been acquired according to the quantied results displayed in the result visualization module; wherein the hierarchical analysis module performs quantitative processing on the conditions for consolidating information in the consolidation condition module, outputs the quantied results through the result visualization module, and determines whether information has been consolidated according to the quantied results displayed in the result visualization module; and wherein the multifaceted Rasch module performs quantitative processing on the conditions for transferring information in the transfer condition module, outputs the quantied results through the result visualization module, and determines whether information has been transferred according to the quantied results displayed in the result visualization module. 2. A basic education enhancement system based on deep learning according to embodiment 1, wherein the SPSS analysis module performs quantitative processing on the conditions required for activating information by using SPSS software; wherein performing quantitative processing on the conditions required for activating information by using the SPSS software specically comprises: an SPSS data preparation stage, utilizing functions provided by the SPSS software to prepare a data le, including dening the structure of SPSS data, and entering and modifying data in a data editing window; an SPSS data processing and organizing stage, performing preprocessing on the data in the data editing window; an SPSS data analysis stage, selecting a correct analysis method to perform data analysis and reverse modeling, wherein the SPSS software is capable of autonomous modeling and automatically performing calculations to provide results; and an analysis result interpretation stage, clarifying the meaning of the calculation results provided, and reasonably interpreting the results in combination with actual background knowledge. 3. A basic education enhancement system based on deep learning according to embodiment 1, wherein the DINA cognitive diagnosis module performs quantitative processing on the conditions required for acquiring information by using a DINA model; wherein the DINA model is expressed as: _ _ In; l-Ii'r'j F(x,-,- _ 1|a,-) (1 s,.) g}. 1 ( ) K _ "Uli Tl.; _ ik k=l Sí = P(X' = Ulm; = 1:] =p X..=1=Û 9} I: u Im} :] (2) wherein the DINA model is expressed as: Xij in which X denotes a score of student i on question j; a,- denotes a mastery status of student i over knowledge points; az,-k denotes a mastery status of student i over knowledge point k; q jk denotes an examination status of question j over knowledge point k; ni]- is equal to 1 or 0, wherein ni]- = 1 indicates that the examinee i has mastered all attributes measured by item j, and nij = 0 indicates that the examinee i has not mastered at least one attribute measured by item j; Sj, referred to as an item slip parameter, denotes a probability that the examinee, although possessing all necessary attributes to answer item j, answers incorrectly due to his or her own error; gj, referred to as an item guess parameter, denotes a probability that the examinee, without fully mastering all necessary attributes to answer item j, answers correctly by chance; to ensure model identiability, for each item P(X = 1 | nij-= 1) Z P(X = 1 | nij-= 0); an accuracy of the model is affected by Sj and gj; wherein performing quantitative processing on the conditions required for acquiring information by using the DINA model specically comprises: constructing an adjacency matrix A and a reachability matrix R; obtaining a complete event matrix Q from the reachability matrix R; obtaining a reduced event matrix Qr from the complete event matrix Q; obtaining a transpose matrix Qr from the reduced event matrix Qr, namely an ideal mastery pattern (IMP); and forming a diagnostic testing framework from results calculated based on the reduced event matrix Qr; constructing the adjacency matrix A according to whether there is a direct logical relationship between each pair of attributes, wherein 1 indicates the presence of a direct relationship and 0 indicates the absence of a direct relationship; constructing the reachability matrix R according to whether there is a direct, indirect, or self-relationship between each pair of attributes, wherein 1 indicates the presence of such a relationship and 0 indicates the absence of such a relationship; obtaining the complete event matrix Q from the R matrix by applying an expansion algorithm to obtain a relationship between attributes and possible sets of item types; wherein the complete event matrix Q contains 26 = 64 possible item sets, deleting items that do not conform to the logical relationships among attributes, and deleting a column that examines zero attributes to obtain the reduced event matrix Qr, which serves as a preparation framework for cognitive diagnostic testing; performing rowcolumn transposition on the reduced event matrix Qr to obtain the transpose matrix Qr, which theoretically reects all attribute mastery patterns of the examinee that conform to the hierarchical relationship of attributes, referred to as the ideal mastery pattern or knowledge state, and adding a row of all- zero matrix to the transpose matrix Qr; encoding student answer data in a data table in binary form, wherein 1 denotes a correct answer and 0 denotes an incorrect answer, and saving the data as a CSV le; and uploading the encoded student answer data and the corresponding Q matrix to a cognitive diagnosis analysis platform, selecting a suitable DINA model, and obtaining the students attribute mastery pattern and attribute mastery probability. 4. A basic education enhancement system based on deep learning according to embodiment 1, wherein the hierarchical analysis module performs quantitative processing on the conditions required for consolidating information by using a hierarchical analysis model; wherein the constructed hierarchical structure model comprises a highest layer, an intermediate layer, and a lowest layer; for any two adjacent layers, an upper layer is referred to as a target layer and a lower layer is referred to as a factor layer; factors in the same layer are subordinate to factors in the upper layer or have an inuence on the factors in the upper layer, and at the same time govern factors in the lower layer or are affected by factors in the lower layer; the highest layer serves as the target layer, being the purpose of solving a problem, and usually contains only one factor; the intermediate layer contains one or more factors, being the criteria that must be followed by various measures or schemes selected to achieve the overall objective, and is also referred to as a strategy layer, a constraint layer, or a criterion layer; when the number of criteria exceeds nine, a sub-criterion layer is irther decomposed; and the lowest layer serves as a scheme or object layer. 5. A basic education enhancement system based on deep learning according to embodiment 4, wherein performing quantitative processing on the conditions required for consolidating information by using the hierarchical analysis model specically comprises: constructing a judgment matrix A, starting from a second layer of the hierarchical structure model; for factors in the same layer that are subordinate to or inuence each factor in the upper layer, constructing the judgment matrix A until reaching the lowest layer; when determining the weights among factors in various layers, adopting a consistent matrix method, namely: not comparing all factors together, but comparing them in pairs; when comparing factors in pairs, adopting a relative scale so as to minimize the difculty of comparing factors of different nature as much as possible and to improve accuracy; wherein the judgment matrix A represents a comparison of the relative importance of all factors in the current layer with respect to one factor in the upper layer, and elements aij of the judgment matrix A are given by using Sattys 19 scale method, namely: an importance intensity of 1 indicates that two factors are of equal importance; an importance intensity of 3 indicates that, compared with another factor, one factor is slightly more important; an importance intensity of 5 indicates that, compared with another factor, one factor is clearly more important; an importance intensity of 7 indicates that, compared with another factor, one factor is more strongly important; an importance intensity of 9 indicates that, compared with another factor, one factor is extremely important; importance intensities of 2, 4, 6, and 8 indicate transitional judgments of importance; an importance intensity given as a reciprocal indicates that if factor i is judged to have importance b compared with factor j, then factor j is judged to have importance 1 / b compared with factor i; wherein the judgment matrix A is expressed as: _ aZJmXn... ... ... ... (4) wherein elements in the matrix A satisfy ai]- > 0, ai]- = ai, aii = 1; and so on, ji until a complete judgment matrix A is established; after the construction of the judgment matrix is completed, performing a hierarchical single ranking and its consistency test; wherein the hierarchical single ranking specically comprises: for a certain element in the upper layer, performing pairwise comparison with all elements in the current layer, and conducting hierarchical ranking to arrange the order of importance; wherein the ranking calculation is performed based on the judgment matrix A, and during the calculation ensuring that the judgment matrix A conforms to eigenvalue and eigenvector conditions, namely: AW=7tmaxW (5) wherein a maximum eigenvalue of the matrix A is Àmax, and a normalized eigenvector of Àmax is W; Wi is a component of W, namely a weight value of the corresponding element in the single ranking; calculating, by using the judgment matrixA, the weights of each factor aij with respect to the target layer, namely calculating a weight vector W and the maximum eigenvalue ).max; wherein the steps for calculating the weight vector W and the maximum eigenvalue )tmax comprise: calculating a product of the elements in each row, and then taking the nth root to obtain an n-dimensional vector, which is expressed by the formula: Wi=nfnÿ=1aij i,j=1. 2. 3......n (6) normalizing Wi so that the sum of all elements in the vector is equal to 1, which is expressed by the formula: Wi (7) Wi _ zi1=1Wi after normalizing Wi, obtaining a ranking weight vector denoted as W; elements in W are ranking weight values representing the relative importance of factors in the same layer with respect to a certain factor in the upper layer; thus W = (W1,W2, Wn)T is the weight vector to be obtained, and is also the result of the hierarchical single ranking sequence of the judgment matrix; wherein the maximum eigenvalue of the judgment matrix A is expressed by the formula: 1 (AW) i Àmaxzz Z?:lTi (8) solving for the maximum eigenvalue and a consistency index (CI) value: assuming that an n-order judgment matrix is B, obtaining its maximum eigenvalue ?.max, which is expressed as: BW=7tW (9) wherein W is an eigenvector of B; in the hierarchical analysis method, a consistency index (CI) is used to test the consistency of judgments, which is expressed as: À n C_L = L n 1 (10) C.I. = 0 indicates that the judgment matrix is completely consistent, and the greater the C.I. value, the more severe the inconsistency of the judgment matrix; based on the CI and RI values, calculating a consistency ratio (CR) value of the hierarchical single ranking to determine whether the judgment matrix A passes the consistency test and whether the judgment matrix needs to be modied, which is expressed as: C.I. C.R. = Rl (1 1) wherein the random consistency index (R.I.) values for matrix orders 113 are as follows: when the matrix order is 1, the R.I. value is 0; when the matrix order is 2, the R.I. value is 0; when the matrix order is 3, the R.I. value is 0.58; when the matrix order is 4, the R.I. value is 0.90; when the matrix order is 5, the R.I. value is 1.12; when the matrix order is 6, the R.I. value is 1.24; when the matrix order is 7, the R.I. value is 1.32; when the matrix order is 8, the R.I. value is 1.41; when the matrix order is 9, the R.I. value is 1.45; when the matrix order is 10, the R.I. value is 1.49; when the matrix order is 11, the R.I. value is 1.51; when the matrix order is 12, the R.I. value is 1.54; when the matrix order is 13, the R.I. value is 1.56; when C.R. < 0.1, the consistency degree of the judgment matrixA is considered to be within an allowable range, and the eigenvector of A is used to perform the weight vector calculation; when C.R. 2 0.1, the judgment matrix A should be modied; wherein the hierarchical total ranking and its consistency test comprise: the hierarchical total ranking refers to calculating the weight values of the relative importance of all factors in a certain layer with respect to the highest layer, and this process is carried out from the highest layer to the lowest layer; assuming that A is the highest layer, including m factors whose weight coefcients for the hierarchical total ranking are al. az . ag, am; assuming that B is an intermediate layer, including n factors whose weight coefcients for the hierarchical single ranking are b}......b,{, then bi=271=1 ajbij is the total ranking of the B layer; assuming that in the B layer, B1, B2, Bn with respect to factor Aj in the upper A layer, j = 1, 2, ..., m have consistency indices of hierarchical ranking CIj and random consistency indices RIj , the consistency ratio (CR) of the hierarchical total ranking is: CR _ alÛÏ1 + (12012 + - ° ' + amCIm _ ZaiCIi _CI alRIl + (1sz + ' ° ' + aRIm _ z at,-RI,- _RI (12) when CR < 0.1, the hierarchical total ranking is considered to have passed the consistency test; otherwise, the element values of the judgment matrix are readjusted, and a nal decision is made based on the hierarchical total ranking of the lowest layer, namely the decision layer. 6. A basic education enhancement system based on deep learning according to embodiment 1, wherein the multifaceted Rasch module performs quantitative processing on the conditions required for transferring information by using a Rasch model; wherein the Rasch model is expressed as: P (Xmi=1l9m 5.)=eXP (Öm-Ö.) / [1+6Xp (Öm-Öi (13) wherein P (Xmi=1|6m 6,) denotes a probability that an individual with ability 9m correctly answers (x = 1) an item with difculty 6,; wherein performing quantitative processing on the conditions required for transferring information by using the Rasch model specically comprises: evaluating, based on data from test results, an item difculty level and a student ability level, and placing the student ability level and the item difculty level on the same equal-interval scale for comparison; wherein reliability is mainly evaluated based on item reliability, examinee reliability, error, and separation indices, and validity is mainly tested based on the t of the data to the model, including analyzing and evaluating unidimensionality, itemexaminee correspondence, and datamodel t indicators such as MNSQ, ZSTD, and pointmeasure correlation (PTMEA); analyzing data results by using Winsteps 4.4.0 analysis software to conduct an overall quality inspection, wherein the overall quality inspection chart reects the overall t as well as reliability and separation issues; converting raw scores into logit values and using a Wright map to place student ability and examinee data on the same scale level, so as to intuitively and concisely match item difculty with student ability, and conducting comparative analysis for different item difculties; testing item t, wherein Winsteps provides two forms of chi-square t indices: Outt MNSQ, being the unweighted mean square residual, and Int MNSQ, being the weighted mean square residual; wherein values of Outt MNSQ and Int MNSQ range from 0 to positive innity, with an ideal value of 1 indicating that the actual data t the model ideally; values greater than 1 (undert) indicate that the variance of the empirical data is greater than expected, and values less than 1 (overt) indicate that the variance of the empirical data is less than expected. 7. A basic education enhancement system based on deep learning according to embodiment 6, wherein processing items with problematic item t comprises: deleting problematic items, revising problematic items, revising scoring criteria, and analyzing inuencing factors causing problems in the items; wherein analyzing inuencing factors causing problems in the items comprises: analyzing aspects such as discrimination and unidimensionality.

Claims

1. A system for improving primary education based on deep learning, characterized by including: a module for external information perception, a sensory memory module, a module for working memory, a module for long-term memory, a module for activation conditions, a module for acquisition conditions, a module for consolidation conditions, a module for transfer conditions, a module for quantitative data processing, and a module for visualization of results; where the quantitative data processing module the contains the following components: an SPSS analysis module, a DINA- cognitive diagnostic module, a hierarchical analysis module and a multi-sided Raschmodule; the module for external information perception, as soon as the attention threshold is reached, enters information into the sensory module memory in; as the information in the sensory memory module is activated by the activation conditions module and resonates with the information in the long-term memory module via pattern recognition and matching processes, the information in the Sensory memory module strengthened and goes directly to the module for long-term memory, which manifests externally as automatic and efficient learning; if the information in the sensory memory module becomes too strong processed or does not match the information in the module for long-term memory, the information appears in the module working memory, where it is processed in depth, consolidated and transferred by the acquisition and consolidation modules / transfer conditions required by deep learning, subsequently corresponds and combines with the information in the module for long-term memory, and to the long-term memory module goes, which manifests externally as the mastery of knowledge by in-depth learning processing by students; when knowledge is retrieved, the information in the module is activated for long-term memory and transferred to the module for working memory by the activation conditions module, which is located externally manifests as students' ability to recall knowledge well; the activation conditions in the activation conditions module include: activation through integration of new and old knowledge, activation by contextual signals, and activation by prediction and open-ended analysis; the conditions for strengthening information in the acquisition conditions module include: a classroom environment with high challenge and low anxiety, in-depth information processing and procedural learning based on declarative knowledge; The in-depth processing of information includes: substantive processing, structural processing and localization of connections; Content processing involves refining concepts using of declarative and procedural definitions; structural processing includes information presentation and organization, and three-layer representations ("macro- symbols"); declarative knowledge includes: pattern recognition, concept acquisition and information visualization; procedural knowledge comprises: analysis of general knowledge, localization and connection of knowledge and feedback from knowledge practice; the conditions for consolidation of information in the consolidation conditions module includes: goal-oriented exercise and varied exercise, spaced exercise and alternating exercise, collaborative task exercise in contexts, concentrated exercise and distributed exercise, physical exercise and mental exercise; the conditions for transfer of information in the transfer conditions module include: model processing, which entails the extracting substantive information from knowledge, finding causal logic between concepts, and the summarizing of concepts to spatial, temporal, causal, logical and quantitative intellectual models to forms; systematic problem solving, which entails the correct describe, analyze, build solutions, evaluate and developing solutions using acquired information, and reasonable interpretation of data from various sources to the to acquire the ability to systematically solve problems; and metacognitive feedback exercise, which involves monitoring someone's performance and thinking to promote someone's behavior and thinking, including feedback on professional knowledge, task knowledge, and strategy knowledge; the SPSS analysis module quantifies the activation conditions of the activation conditions module, and the results are displayed via the module for visualization of results, in which it is assessed whether the information is activated based on the quantitative results in the module for visualization of results are displayed; The DINA cognition diagnostic module quantifies the acquisition conditions of the acquisition conditions module, and the results are displayed via the visualization module results, whereby it is assessed whether the information was acquired on the basis of the quantitative results in the module for visualization of results are displayed; the hierarchical analysis module quantifies the consolidation conditions of the consolidation conditions module, and the results are displayed via the visualization module of results, whereby it is assessed whether the information is consolidated on based on the quantitative results in the visualization module of results are displayed; The multi-sided Rasch module quantifies the transfer conditions of the transfer conditions module, and the results are displayed via the results visualization module, where assessment takes place whether the information has been transferred based on the quantitative results which are displayed in the Results Visualization module.

2. The deep learning-based system for improving the primary education according to conclusion l, characterized because: the SPSS analysis module the conditions quantifies those needed to activate information using of SPSS software; the quantification of the conditions required for activating information using SPSS software specifically includes: the data preparation phase, where the functions that by SPSS offered are used to data files for prepare, including defining the structure of SPSS data in the data editing window, and entering and modifying facts; the data processing and organization phase, where data in the data editing window is pre-processed; the data analysis phase, in which the correct analysis method is selected for data and reverse analysis modeling, and SPSS- software can model autonomously and automatically the results calculate; the interpretation phase of analysis results, where the meaning of the calculated results are clarified and become reasonable interpreted in combination with factual background knowledge.

3. The deep learning-based system for improving the primary education according to conclusion ], characterized by: the DINA- cognition diagnostic module quantifies the conditions necessary for the acquiring information using the DINA model; the DINA model is displayed as: He I-HÙ' PX--=la-= 13- . [ UI l') ( J.) 9; (1} .tr _ m me! to: 1 s]- = FW:; = Im; = 1) g, = F(x,-_, = ilm, = ü) Where Xij refers to student i's score on item j, ai refers to to the mastery of the knowledge by the student, aik refers to the mastery of knowledge point k by the student, and qjk refers to the extent to which item j the knowledge center k assesses; Tlij is equal to 1 or 0; my = 1 indicates that the examined person i masters all attributes tested by item j, while nij = 0 indicates that the investigated person has at least one of the attributes which is tested by item j, has not mastered; Sj becomes the item- called the slip parameter, which indicates the probability that the examined person, although he possesses the necessary attributes to solve item ], the item answers incorrectly due to its own errors; gj is called the item guessing parameter, which the probability indicates that the investigated person, who the necessary attributes to solve item j not fully mastered, the item answers correctly by chance; to the identifiability of the to safeguard the model, it must hold for each item that P( Xij =1 Im,- =1) z P( Xi] =1 Inij =O); the accuracy of the model is affected by Sj and g;. The quantitative analysis of the conditions required for the Acquiring information using the DINA model includes specifically the following steps: Constructing the proximity matrix A and the range matrix R; the obtaining the complete event matrix Q based on the range matrix R; deriving the reduced event matrix Qr from the complete event matrix Q; obtaining the transposed matrix Qr' from the reduced event matrix Qr, which theoretically the ideal control patterns of the investigated persons according to the represents the hierarchical relationship of attributes, also known as knowledge states mentioned, and adding a row of zeros to the transposed matrix Qr'; Coding the students' response data in the dataset with 0-lcoding, where correct answers are coded as 1, and incorrect answers are coded as 0, and then stored as a CSV file; Uploading the encoded response data together with the corresponding Q-matrix to the cognitive diagnostic analysis platform, selecting the correct DINA model, and obtaining the mastery patterns and probabilities of the students.

4. The deep learning-based system for improving the primary education according to conclusion l, characterized by the hierarchical analysis module the quantitative analysis of the necessary conditions for consolidating information performs using the hierarchical analysis model; the constructed hierarchical The structural model comprises the highest level, the middle level, and the lowest level. For two adjacent levels, the higher level becomes called the target level and the lower level the factor level. The factors belong to or influence the factors in the higher level at the same level level, and master simultaneously or be influenced by the factors in the lower level. The highest level is the target level, which usually consists of only one factor exists, which is the goal to solve the problem; the The middle level consists of one or more criteria that must be followed. are to achieve the general goal and are also known as the mentioned strategic, constraint, or criterion level; if there are more than 9 if criteria are, these can be further subdivided into sub-criteria levels; The lowest level is the alternatives or objects level.

5. The deep learning-based system for improving the primary education according to conclusion 4, characterized by the quantitative analysis of the conditions required for consolidating information specifically using the hierarchical analysis model includes: Constructing the judgment matrix A, starting from the second level of the hierarchical structural model, constructing the judgment matrix A for the factors at the same level that belong to or influence every factor at the higher level, and this process continue until the lowest level is reached; When determining the weights of the factors at each level, is uses the consistent matrix method, which includes: Do not compare all factors at the same time, but instead comparing pairwise; The use of a relative scale for pairwise comparisons of factors to the difficulty of the to minimize comparing factors of different nature and the to improve accuracy; The judgment matrix A represents the pairwise comparison of the relative importance of all factors in the current level compared to of a specific factor at the higher level, and the elements ai] of the Judgment matrix A is provided using the 1-9 scale method. of Saaty, specifically: A value of 1 indicates that the two factors to be compared are equal. be important; A value of 3 indicates that a factor something is more important than the other; A value of 5 indicates that one factor is clearly more important. than the other; A value of 7 indicates that a factor is significant is more important than the other; A value of 9 indicates that one factor is extremely more important than the others; Values ​​of 2, 4, 6, and 8 indicate intermediate levels of importance to the relationship between the main values; Reciprocal values ​​are used to describe the relationship between factor i and to express factor j, where the equation of factor i with factor j is is given if b, the equation of factor j with factor i becomes given as l / b; The judgment matrix A is displayed as: . O . mxn . 0 OO 0 0 0 O 0 O . . (4) Whereby the elements in A satisfy 1 H ?Ü: H = ;: aii=1 '_Il ; this process continues until a full Judgment matrix A has been established. After drawing up the judgment matrix, becomes a hierarchical single ranking and consistency check performed. The hierarchical simple arrangement comprises pairwise comparisons between an element of the previous level and all elements of the current level, followed by hierarchical sorting to the to determine order of importance. The ranking calculations are based on the judgment matrix A, ensuring that the matrix meets the conditions for eigenvalue and eigenvector, specific: AW=AmaxW ( 5 ) Where Xmax is the maximum eigenvalue of matrix A, and W the is the normalized eigenvector of Àmax. Wi is the component of W, which the weight of the corresponding element in the single ranking represents. To calculate the weight of each factor ai]- in the target level using the judgment matrix A, the steps for calculating are of the weight vector W and the maximum eigenvalue Xmax as follows: First calculate the product of the elements in each row and take next, the n-th root of the product to an n-dimensional vector to obtain, which is displayed as: î * _ . . _ [[Fla lui1 21 3......n (6) Next, normalize Wi such that the sum of the elements in the vector is equal to 1, which is represented as: W- wi = _l Egg=1wi (7) After normalizing Wi, the sorted weight vector W becomes obtained, where the elements of W the relative importance of the factors at the same level relative to a certain factor in the represent previous level. The sorted weight vector W is then displayed as W = (W1, W2, Wn)T , which is the determine the weight vector and the result of the hierarchical simple ranking of the judgment matrix. The maximum eigenvalue of the judgment matrix A is represented as: _1 n (AW) . lllnii-::.îu:_;í:i=1 W- I ' (8) To calculate the maximum eigenvalue and the consistency index CI: Given a Nord judgment matrix B, the maximum eigenvalue Xmax is obtained, displayed as: BW=AW (9) Where W is the eigenvector of B. In the analytical hierarchy process (AHP), the consistency index CI is used to test the consistency of the judgment, shown as: CI _ AIM7 ' ' n 1 ( 10) Cl. = 0 indicates that the judgment matrix is ​​perfectly consistent; the higher the Cl., the more serious the inconsistency in the judgment matrix. The consistency ratio CR is calculated based on the CI values. and RI to determine whether judgment matrix A has the consistency test pass and whether the judgment matrix needs to be corrected, displayed as: GI CR = RI (11) The random consistency index RI values ​​for matrices of order 1-13 are as follows: For a matrix of order 1, R1. = 0; For a matrix of order 2, Rl = 0; For a matrix of order 3, RI = 0.58; For a matrix of order 4, RI = 0.90; For a matrix of order 5, R.

1. = 1.12; For a matrix of order 6, R.

1. = 1.24; For a matrix of order 7, R.

1. = 1.32; For a matrix of order 8, RI = 1.41; For a matrix of order 9, RI = 1.45; For a matrix of order 10, R.

1. = 1.49; For a matrix of order 11, R.

1. = 1.51; For a matrix of order 12, RI = 1.54; For a matrix of order 13, RI = 1.56; When CR < 0.1, the consistency of the judgment matrix A is determined as considered acceptable, and the weight calculation can be continued with the eigenvector of A; when CR 2 0.1, the judgment matrix A must be revised. Hierarchical overall ranking and the consistency test: Hierarchical total ranking refers to the calculation of the weights of all factors at a certain level relative to the most important factor at the highest level. This process proceeds from the highest to lowest level: Let A be the highest level, including m factors with total ranking weights al , az , a3, am ; let B be the middle be level, including 11 factors with single ranking weights bi.-....ba. hama-bi,- _ _ _ the total ranking of the B-level; Let the consistency index for the hierarchical ranking of the B- level relative to factor Aj in the higher level CI j be, and the random consistency index RI j, then the consistency ratio CR for the total ranking is: CR _ (11011 + (12012 + ' ' ' + aCIm _ 204015 _CI alRI1 + agRI2 + - ' ' + amRIm E (z,-RI,- RI ( 12) When CR < 0.1, the total ranking is deemed to be to have passed the consistency test; otherwise the element values ​​must of the judgment matrix are adjusted. Based on the total ranking of the lowest level, i.e. the decision-making level, the final decision is made.

6. The deep learning-based system for improving the primary education according to conclusion 1, characterized by the multifaceted Rasch module the quantitative analysis of conditions that are necessary for information transfer performs using the Rasch model; The Rasch model is expressed as: P (X mi=1|9 ö,)=exp (aay [1+exp (B-&)] ( 13 ) P (X mi=1 lam, 6i) refers to the probability that an individual with ability H correctly answers an item with difficulty 6.- when "x=1." The Rasch model quantitatively analyzes the conditions that are necessary is for information transfer through item difficulty levels and 1 to evaluate student skill levels based on test results. The places the skill levels of the students and the difficulty levels of the items on the same interval scale for comparison. Reliability is primarily evaluated based on of item reliability, candidate reliability, error and separation index, while validity is primarily assessed by the examining the data model t, including unidimensionality, item candidate correspondence, data model t indices such as MNSQ, ZSTD, and point measure correlation PTMEA. Data results are analyzed using Winsteps 4.4.0 software that performs overall quality tests. The Greek for overall quality tests reflect the overall t, reliability and separation problems. Raw values ​​are converted into logit values, and a Wright card is used to assess the student's skill and to place candidate data on the same scale, creating an intuitive and simple matching and comparison of item difficulty and student skill levels become possible, as well as a comparative analysis of different item difficulties. Winsteps offers two types of chi-square t indices to the item tests: Outt MNSQ, that the unweighted mean square residual represents, and Int MNSQ, that the weighted average represents quadratic residue. The value range for Outt MN SQ and Int MNSQ range from 0 to positive infinity, with an ideal value of 1, which indicates that the actual data fit well with the model. Values ​​greater than 1 indicate undert, which means that there is more variation is then expected in the empirical data, while l values ​​smaller Less than 1 indicates overt, which means there is less variation than expected. in the empirical data.

7. The deep learning-based system for improving the primary education according to conclusion 6, characterized by dealing with items with problematic t include: removing, revising problematic items, revision of assessment