Teaching quality assessment methods and systems based on general education

Abstract

This invention relates to the field of teaching evaluation technology, specifically to a teaching quality evaluation method and system based on general education. The method includes the following steps: extracting a sequence of ability achievement and constructing a stable interval; removing outliers to form discrete nodes; calculating course difficulty indicators and difficulty compensation scores; identifying effective continuous scoring periods and generating segmented score averages; and inputting each result into a support vector regression model to generate a quality evaluation result. In this invention, by jointly correcting for abnormal fluctuations in course achievement difficulty and changes in evaluation rhythm, the interference of outlier data on the overall conclusion is reduced, enhancing the comparability between different courses and different stages, improving the stability and interpretability of the evaluation results, making quality judgments closer to the actual teaching situation, and increasing the reference and application value of the evaluation results in teaching management, providing a more targeted basis for teaching feedback and continuous improvement.

Description

Technical Field

[0001] This invention relates to the field of teaching assessment technology, and in particular to a method and system for assessing teaching quality based on general education. Background Technology

[0002] The field of teaching evaluation technology encompasses methods and systems for measuring, analyzing, and judging the entire teaching process. It mainly involves core matters such as setting course objectives, recording the learning process, observing teaching behaviors, measuring learning outcomes, constructing evaluation indicators, and providing feedback on evaluation results. This field typically conducts a comprehensive multi-dimensional judgment on teaching quality by systematically analyzing the relationship between the implementation process of teaching content and learning output, combining quantitative data and qualitative descriptions. It also includes the formulation of evaluation standards, the collection of evaluation data, the analysis of evaluation results, and continuous improvement mechanisms, thereby forming a relatively complete teaching quality monitoring and evaluation system.

[0003] Traditional teaching quality assessment methods refer to a type of assessment approach that focuses on judging teaching effectiveness. It mainly targets the correspondence between the course teaching process and learning outcomes. By setting a unified evaluation index system, it collects specific data such as classroom teaching performance, homework completion, exam scores, and questionnaire feedback. The scores are calculated and weighted according to preset scoring rules. Typically, a manually designed scoring table is used to record the teacher's teaching performance, student participation, and assessment results item by item. The final evaluation conclusion is formed by statistically analyzing the scores of each item. At the same time, horizontal comparisons are made between different courses or different teachers according to a predetermined evaluation cycle to complete the processing of relevant technical matters for teaching quality assessment.

[0004] Traditional assessment methods rely on uniform indicators and fixed periods for score aggregation. The assessment process focuses on the static collection of classroom performance, homework scores, and questionnaire results, which makes it difficult to reflect the real relationship between differences in ability achievement, fluctuations in course difficulty, and continuous changes in scores in general education. Horizontal comparison conclusions are easily affected by abnormal data. The evaluation results are insufficient in identifying the stage characteristics of the teaching process, resulting in weak stability of quality judgment. The feedback content is relatively general, and the basis for continuous improvement is not specific enough, making it difficult to support the refined assessment of the quality of general education teaching. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a teaching quality assessment method and system based on general education.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a teaching quality assessment method based on general education, comprising the following steps: S1: Extract the sequence of general education ability achievement, calculate the median value of each ability and the difference between each ability and the median value, construct the median difference sequence, sort and select the set of values ​​within the preset ratio range on both sides as the stable candidate set, and calculate the mean and standard deviation to generate the stable interval mean and standard deviation range. S2: Based on the mean of the stable interval and the standard offset range, outlier judgment is performed on the capability achievement sequence, values ​​exceeding the outlier judgment threshold are removed, and the remaining achievement data are divided into equal intervals and the interval mean is calculated to construct discrete nodes of capability achievement. S3: Extract raw grade data from the general education course assessment database, and combine it with the average course grade, course difficulty index, preset benchmark difficulty and difficulty offset to generate a difficulty-compensated grade sequence. S4: Collect continuous scoring data sequences through the general education evaluation portal, construct scoring change sequences, screen continuous stable segments and select effective period candidate sets, and generate evaluation period parameters; S5: Calculate the average score of each segment based on the evaluation period parameters, and input the discrete nodes of the ability achievement, the difficulty compensation score sequence and the average score of each segment into the support vector regression model to perform regression estimation operation and generate the general education quality assessment results.

[0007] The present invention is improved in that the step of obtaining the standard offset range is specifically as follows: S111: Obtain general education records, extract the general ability achievement sequence, sort each element in the general ability achievement sequence in ascending order of value, extract the center value corresponding to the sorting node in the queue, obtain the median value of general ability, calculate the difference between each element in the general ability achievement sequence and the median value of general ability, and establish a median value difference sequence. S112: Perform ascending sorting on the median difference sequence to construct an ascending difference sequence, obtain a preset bilateral boundary ratio, calculate the single-sided truncation number based on the preset bilateral boundary ratio, truncate and remove extreme values ​​on the ascending difference sequence based on the single-sided truncation number, extract the remaining middle position values, and construct a stable candidate set. S113: Call each value in the stable candidate set to perform a summation operation, divide the summation result by the total number of elements covered by the stable candidate set to obtain the mean of the stable interval, calculate the sum of squares of the difference between each value in the stable candidate set and the mean of the stable interval, divide the sum of squares by the total number of elements covered by the stable candidate set to perform a square root operation, and calculate the standard offset range.

[0008] The present invention is improved in that the step of obtaining the discrete nodes of the capability achievement degree is specifically as follows: S211: Call the capability achievement sequence and the mean of the stable interval, perform loop traversal extraction for each element of the capability achievement sequence, calculate the absolute value of the difference between each element and the mean of the stable interval, perform numerical aggregation processing on all extracted absolute values, and construct the achievement deviation set. S212: Based on the achievement deviation set, extract the value of each deviation, obtain the preset offset multiple and the standard offset range, calculate the corresponding product value of the preset offset multiple and the standard offset range to construct the outlier determination threshold, compare each deviation value with the outlier determination threshold, locate the associated value of the capability achievement sequence when the deviation value is greater than the outlier determination threshold, remove the associated value from the capability achievement sequence, and merge the retained items to generate the remaining achievement data; S213: Rearrange the remaining achievement data in ascending order to obtain an ascending achievement sequence, obtain a preset interval quantity parameter, divide the total number of elements in the ascending achievement sequence by the preset interval quantity parameter to calculate the single interval carrying quota, perform equal-quantity cutting on the ascending achievement sequence according to the single interval carrying quota, construct an equal-quantity data segment set, calculate the mean value of each segment in the equal-quantity data segment set, and summarize to establish discrete nodes of capability achievement.

[0009] The present invention is improved in that the offset multiple is set based on obtaining a historical ability achievement sample set of the same type of general education courses, calculating the cumulative probability density distribution data of the historical ability achievement sample set under multiple standard deviation spans, combining the preset effective sample retention ratio, searching the distribution node in the cumulative probability density distribution data that matches the effective sample retention ratio, and extracting the standard deviation multiplier corresponding to the distribution node as the offset multiple.

[0010] The present invention is improved in that the steps for obtaining the difficulty compensation score sequence are as follows: S311: Extract raw score data from the general education course assessment database, perform a summation operation on all the scores in the raw score data, divide the summation value by the total number of students to obtain the course average score, calculate the difference between the raw score data and the course average score, perform a square operation on each difference to construct a set of score square differences, add all the elements in the score square difference set and divide by the total number of students to obtain the course difficulty index. S312: Obtain a preset benchmark difficulty index, perform a subtraction calculation based on the course difficulty index and the preset benchmark difficulty index to obtain the difficulty difference, extract the absolute value of the difficulty difference as the absolute value of the offset, determine the positive or negative attribute of the value of the difficulty difference to extract the offset direction factor, and perform a multiplication operation between the absolute value of the offset and the offset direction factor to generate the difficulty offset. S313: The original score data and the average course score are used to calculate the difference to obtain the individual score deviation. The preset difficulty reverse weighting parameter is extracted. The individual score deviation is multiplied by the difficulty offset and the difficulty reverse weighting parameter to calculate the score correction increment. The score correction increment is superimposed on the original score data to obtain the global correction score set. The global correction score set is reorganized and sorted according to the personnel identity to establish the difficulty compensation score sequence.

[0011] The present invention is improved in that the steps for obtaining the evaluation period parameters are as follows: S411: Collect continuous scoring data sequences through the general education evaluation portal, perform subtraction operations on adjacent node values ​​within the continuous scoring data sequence to construct a scoring change sequence, extract all differences within the scoring change sequence and sum them to obtain the total difference, divide the total difference by the total number of elements in the scoring change sequence to calculate the mean, and establish a scoring change reference benchmark. S412: Obtain a preset judgment ratio factor, multiply the score change reference benchmark by the preset judgment ratio factor to calculate the fluctuation judgment threshold, extract the segments in the score change sequence that are continuously lower than the fluctuation judgment threshold to form a continuous stable segment, read the start and end timestamps of the continuous stable segment and subtract them to obtain the duration span, retain the continuous stable segments associated with the duration span greater than the preset length judgment benchmark, and establish an effective period candidate set. S413: For the effective period candidate set, extract the highest score and the lowest score in each period and perform subtraction to obtain the interval fluctuation range. Divide the interval fluctuation range by the corresponding duration span to calculate the fluctuation ratio coefficient. Sort all the calculated fluctuation ratio coefficients in ascending order and extract the first value of the queue as the minimum value item. Read the period length data mapped by the minimum value item to generate the evaluation period parameter.

[0012] The present invention improves upon this by setting the determination ratio factor based on the background fluctuation characteristics of the historical evaluation dataset within a known steady-state teaching period. By extracting multiple sets of historical steady-state evaluation sample sequences, the ratio of the local score dispersion to the global score mean change within each set sequence is calculated to construct a set of background error ratio distributions. Based on a statistical probability density model, the convergence upper bound critical coefficient value of the set of background error ratio distributions under a pre-set reliability condition is extracted. The critical coefficient value is used as the maximum tolerance scale limit for the fluctuation of the true evaluation after excluding the interference of normalized random scoring, and is established as the determination ratio factor.

[0013] The present invention is improved in that the steps for obtaining the general education quality assessment results are specifically as follows: S511: Perform equidistant segmentation on the continuous scoring data sequence according to the evaluation cycle parameters to construct a set of scoring sub-segments. Extract the scoring values ​​covered by each sub-segment of the scoring sub-segment set and perform an accumulation operation to obtain the sub-segment sum. Divide the sub-segment sum by the number of elements in the corresponding sub-segment to calculate the average value and generate the segmented scoring mean. S512: The discrete nodes of the ability achievement degree, the difficulty compensation score sequence and the average of the segmented scores are column-wise concatenated according to the preset index to extract a multi-source feature matrix. Each element in the multi-source feature matrix is ​​divided by a preset extreme value constant to perform normalization mapping, and the numerical values ​​are compressed to extract a multi-dimensional evaluation feature vector. S513: Call the preset support vector regression model, import the multidimensional evaluation feature vector into the support vector regression model, call the kernel function associated with the support vector regression model to calculate the inner product value of the multidimensional evaluation feature vector and the support vector, multiply the inner product value by the corresponding weight coefficient and add the bias constant to perform regression estimation operation, extract the estimated value to generate the general education quality assessment result.

[0014] The present invention improves upon the previous method by multiplying the inner product value by the corresponding weight coefficient and adding the bias constant to perform regression estimation. Specifically, the process involves: determining the kernel function associated with the support vector regression model as a Gaussian function; minimizing the error of the historical evaluation vector set by the support vector regression model; calculating the square of the absolute value of the difference between the multidimensional evaluation feature vector and the support vector to obtain the squared distance value; obtaining the kernel function parameter; multiplying the squared distance value by the negative of the kernel function parameter to obtain the decay value; raising the natural base to the power of the decay value to obtain the inner product value; setting the kernel function parameter based on the reciprocal of the support vector distribution distance; reading the corresponding weight coefficient and bias constant; setting the corresponding weight coefficient according to the difference of the penalty factor in the optimization solution; setting the bias constant based on the average difference between the actual output of the support vector and the expected inner product value; multiplying the inner product value by the corresponding weight coefficient and summing the results to obtain the comprehensive mapping value; and performing an addition operation on the comprehensive mapping value and the bias constant to obtain the estimated value.

[0015] A teaching quality assessment system based on general education, wherein the teaching quality assessment system based on general education is used to implement the above-mentioned teaching quality assessment method based on general education, the system comprising: The benchmark screening module extracts the sequence of general education ability achievement, calculates the median value of the ability and the difference between each item and the median value of the ability, constructs the median difference sequence, sorts and selects the set of values ​​within the preset ratio range on both sides as the stable candidate set, and calculates the mean and standard deviation to generate the mean and standard deviation range of the stable interval. The node integration module performs outlier detection on the capability achievement sequence based on the mean of the stable interval and the standard offset range, removes values ​​that exceed the outlier detection threshold, performs equal interval division on the remaining achievement data and calculates the interval mean, and constructs discrete nodes of capability achievement. The grade correction module extracts raw grade data from the general education course assessment database and combines it with the course average grade, course difficulty index, preset benchmark difficulty and difficulty offset to generate a difficulty compensation grade sequence. The period identification module collects continuous scoring data sequences through the general education evaluation portal, constructs scoring change sequences, filters continuous stable segments and selects a set of valid period candidates, and generates evaluation period parameters. The quality assessment module calculates the average score for each segment based on the evaluation period parameters, and inputs the discrete nodes of the ability achievement, the difficulty compensation score sequence, and the average score for each segment into the support vector regression model to perform regression estimation and generate the general education quality assessment results.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, a stable interval is first constructed around the competency achievement data, and outlier removal is completed. Discrete nodes are formed for the remaining data. Difficulty compensation results are generated by combining the relationship between the original scores and the average scores. After identifying the effective period based on continuous score changes, segmented mean processing is implemented. This allows the difficulty of competency assessment and the pace of teaching evaluation to be collaboratively corrected and integrated within the same evaluation link, reducing the influence of abnormal fluctuations on the overall conclusion, enhancing the comparability between different courses and different stages, making the output results closer to the real teaching situation, providing more targeted basis for teaching feedback and continuous improvement, and improving the stability, interpretability, and application value of the evaluation conclusions. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Please see Figure 1 This invention provides a technical solution, a teaching quality assessment method based on general education, comprising the following steps: S1: Extract the general education competency achievement sequence through the general education academic affairs system, calculate the median value of the competency achievement sequence, and construct a median value difference sequence by calculating the difference between each item in the competency achievement sequence and the median value. Sort the median value difference sequence, select the set of values ​​within the preset ratio range on both sides as the stable candidate set, calculate the mean and standard deviation of the stable candidate set, and generate the mean and standard deviation range of the stable interval. S2: Based on the mean of the stable interval and the standard offset range, outlier judgment is performed on the capability achievement sequence. When the difference between a specific achievement value and the mean of the stable interval exceeds a preset multiple of the standard offset range, the corresponding achievement value is removed from the capability achievement sequence to obtain the remaining achievement data. The remaining achievement data is divided into equal intervals, and the mean of each interval is calculated to construct discrete nodes of capability achievement. S3: Extract raw score data from the general education course assessment database and calculate the average course score. Square the difference between the raw score data and the average course score and calculate the mean to generate a course difficulty index. Calculate the difference between the course difficulty index and the preset benchmark difficulty to generate a difficulty offset. Use the raw score data, the average course score and the difficulty offset to generate a difficulty compensation score sequence. S4: Collect continuous scoring data sequences through the general education evaluation portal, perform adjacent numerical difference operations on the continuous scoring data sequences to construct scoring change sequences, calculate the overall average change value of the scoring change sequences, compare each item of the scoring change sequences with the overall average change value, select multiple time segments with continuous differences less than a preset proportion of the overall average change value as continuous stable segments, count the time span of continuous stable segments, select continuous stable segments that meet the minimum continuous length threshold to form an effective period candidate set, calculate the fluctuation amplitude of the effective period candidate set and perform a ratio operation with the current span, select the period length corresponding to the minimum ratio value to generate evaluation period parameters; S5: Based on the evaluation period parameters, the segmented average is calculated on the continuous scoring data sequence to construct the segmented average score. The discrete nodes of ability achievement, the difficulty compensation score sequence and the segmented average score are input into the support vector regression model for fusion calculation to generate the general education quality assessment results.

[0020] The specific steps for obtaining the standard offset range are as follows: S111: Obtain general education records, extract the general ability achievement sequence, sort each element in the general ability achievement sequence in ascending order of value, extract the center value corresponding to the sorting node in the queue, obtain the median value of general ability, calculate the difference between each element in the general ability achievement sequence and the median value of general ability, and establish a median value difference sequence. The general education records are exported from the academic affairs management platform. These records refer to the original data table containing student identification, course number, and ability achievement values ​​for the same general education course. The export method involves filtering by course number and semester number on the academic affairs management platform and downloading it as a structured table. Then, in Excel or Python, the system checks for missing, duplicate, and blank entries by student identification, deleting blank records and retaining valid ability achievement values. Next, all ability achievement values ​​are read sequentially and written into a general education ability achievement sequence according to the original record order. All values ​​in the sequence are then rearranged in ascending order. Finally, the value in the middle of the sorted sequence is used as the center value, ensuring the total number of elements in the sequence is odd. The system directly reads the value at the middle position. When the total number of elements in the sequence is even, it reads the values ​​at the two middle positions and calculates the mean. The median value of general knowledge ability is calculated as (the value at the beginning of the middle position + the value at the end of the middle position) / 2. Then, it reads each value in the original general knowledge ability achievement sequence one by one and calculates the deviation with the median value of general knowledge ability. The median value difference is calculated as the current ability achievement value minus the median value of general knowledge ability. When the current ability achievement value is greater than the median value of general knowledge ability, it is recorded as a positive difference value. When the current ability achievement value is equal to the median value of general knowledge ability, it is recorded as 0. When the current ability achievement value is less than the median value of general knowledge ability, it is recorded as a negative difference value. Then, it writes all the differences into the median value difference sequence in the original order for direct use when selecting stable intervals later.

[0021] S112: Sort the median difference sequence in ascending order, construct the ascending difference sequence, obtain the preset bilateral boundary ratio, calculate the unilateral truncation number based on the preset bilateral boundary ratio, truncate the ascending difference sequence based on the unilateral truncation number to remove extreme values, extract the remaining middle position values, and construct a stable candidate set. Read all differences from the median difference sequence and re-sort them in ascending order in Excel or Python to obtain an ascending difference sequence. Then, read the preset two-sided boundary ratio. The preset two-sided boundary ratio refers to the proportion of each side of the total number of elements when removing data from both ends of the difference sequence. This ratio is pre-calculated using historical achievement data tables of similar general education courses. During the statistical analysis, historical difference sequences are first summarized by course category, and then the first and last edges are removed sequentially according to different removal ratios. The stability of the remaining data after removal is recorded. The ratio with the smallest fluctuation in the remaining quantity among consecutive batches of data, where the remaining quantity in each batch is greater than 50% of the total quantity, is written into a parameter table. A ratio of 0 indicates that boundary values ​​are not removed, and a ratio greater than 0 and less than 0.5 indicates that boundary values ​​are removed. Both sides are truncated. Then, the total number of elements in the ascending difference sequence is read. The truncation number on one side = the total number of elements in the ascending difference sequence × the preset bilateral boundary ratio. The result is then rounded down to obtain the actual number of elements removed on each side. Next, the smallest difference corresponding to the truncation number is deleted from the beginning of the ascending difference sequence, and the largest difference corresponding to the truncation number is deleted from the end of the ascending difference sequence. If the truncation number on one side is 0, neither the beginning nor the end is deleted. If the number of remaining elements after deletion is greater than 0, all the values ​​retained in the middle are written into the stable candidate set. If the number of remaining elements after deletion is 0, the preset bilateral boundary ratio is reduced to half of the original ratio, the truncation number on one side is recalculated, and the deletion at the beginning and end is repeated until the number of remaining elements is greater than 0.

[0022] S113: Call each value in the stable candidate set to perform a summation operation, divide the summation result by the total number of elements covered by the stable candidate set to obtain the mean of the stable interval, calculate the sum of squares of the difference between each value in the stable candidate set and the mean of the stable interval, divide the sum of squares by the total number of elements covered by the stable candidate set to perform a square root operation, and calculate the standard offset range. Read all values ​​in the stable candidate set one by one and calculate the sum. The mean of the stable interval = the sum of all values ​​in the stable candidate set / the total number of elements in the stable candidate set. Then, read each value in the stable candidate set one by one again, calculate the difference between it and the mean of the stable interval, and then square the difference. Then, sum all the individual squared deviations to obtain the total squared deviation; after that, When the total number of elements in the stable candidate set is equal to 1, the mean of the stable interval is directly taken as this unique value, and the standard offset range is recorded as 0. When the total number of elements in the stable candidate set is greater than 1, the mean and standard offset range are calculated in the aforementioned order, and then the mean of the stable interval and the standard offset range are written into the subsequent achievement deviation comparison table for discrete judgment of the achievement data of each student in the same general education course.

[0023] The specific steps for obtaining discrete nodes of capability achievement are as follows: S211: Call the capability achievement sequence and the mean of the stable interval, perform loop traversal extraction for each element of the capability achievement sequence, calculate the absolute value of the difference between each element and the mean of the stable interval, perform numerical aggregation processing on all extracted absolute values, and construct the achievement deviation set. The system reads the original ability achievement sequence and the mean of the stable interval. The ability achievement sequence refers to the set of original ability achievement values ​​arranged in order of student identity, and the mean of the stable interval refers to the level of central deviation formed by the previously retained data. Then, it iterates through each value in the ability achievement sequence, first calculating the difference between the value and the mean of the stable interval, and then reading the absolute value of the difference to obtain the current achievement deviation: Achievement deviation = |Current ability achievement value - Mean of stable interval|. After that, each achievement deviation is written into the achievement deviation set in the same order as the original ability achievement sequence. When the current ability achievement value is greater than the mean of the stable interval, its positive deviation from the mean is retained; when the current ability achievement value is equal to the mean of the stable interval, the deviation is recorded as 0; when the current ability achievement value is less than the mean of the stable interval, its negative deviation from the mean is retained. This allows subsequent threshold comparisons to be completed directly based on the magnitude of the deviation, without distinguishing between positive and negative directions.

[0024] S212: Based on the achievement deviation set, extract the value of each deviation, obtain the preset offset multiple and standard offset range, calculate the corresponding product value of the preset offset multiple and standard offset range to construct the outlier determination threshold, compare each deviation value with the outlier determination threshold, locate the associated value of the capability achievement sequence when the deviation value is greater than the outlier determination threshold, remove the associated value from the capability achievement sequence, and merge the retained items to generate the remaining achievement data; The offset factor is set based on obtaining a historical ability achievement sample set of similar general education courses, calculating the cumulative probability density distribution data of the historical ability achievement sample set under multiple standard deviation spans, combining the preset effective sample retention ratio, searching for distribution nodes that match the effective sample retention ratio in the cumulative probability density distribution data, and extracting the standard deviation multiplier corresponding to the distribution node as the offset factor. Read the deviation value of each item in the achievement deviation set, and read the preset offset multiple and standard offset range. The outlier determination threshold = preset offset multiple × standard offset range. The preset offset multiple refers to the standard deviation span multiple determined in the current general education course's historical ability achievement samples of the same type to meet the effective sample retention ratio. When setting it, first export the historical ability achievement sample set of the same type of general education courses from the academic affairs management platform, and then summarize it in Python by course category, semester, and grade. Calculate the mean and standard deviation of each batch of samples in batches. Then, take multiple standard deviation spans at 0.1 intervals, and count the proportion of historical samples within the range of "mean ± current span × standard deviation" for each span to obtain the cumulative probability density distribution data. Subsequently, read the preset effective sample retention ratio, which refers to the minimum proportion of historical samples that need to be retained after outlier removal. Its value is obtained through historical... In the general education course data table, the median level of the final valid record count relative to the original record count for each batch is determined. If the cumulative percentage corresponding to a certain standard deviation span is equal to the retention ratio, then that span is directly written as the offset multiple. If the cumulative percentage is greater than the retention ratio for the first time, then that span is written as the offset multiple. If the cumulative percentage corresponding to all spans is less than the retention ratio, then the largest span is taken as the offset multiple. Then, the deviation values ​​in the achievement deviation set are compared with the outlier determination threshold item by item. When the deviation value is greater than the outlier determination threshold, the original ability achievement value corresponding to the deviation is marked as a rejection item. When the deviation value is equal to the outlier determination threshold, the original ability achievement value corresponding to the deviation is marked as a retention item. When the deviation value is less than the outlier determination threshold, the original ability achievement value corresponding to the deviation is marked as a retention item. Finally, all rejection items are deleted, and all retention items are merged in their original order to form the remaining achievement data.

[0025] S213: Rearrange the remaining achievement data in ascending order to obtain the ascending achievement sequence, obtain the preset interval quantity parameter, divide the total number of elements in the ascending achievement sequence by the preset interval quantity parameter to calculate the single interval carrying quota, perform equal division on the ascending achievement sequence according to the single interval carrying quota, construct an equal data segment set, calculate the mean value of each segment in the equal data segment set, and summarize to establish discrete nodes of capacity achievement. Read the remaining achievement data and rearrange it in ascending order to obtain an ascending achievement sequence. Then, read the preset interval quantity parameter, which specifies how many equal segments to divide the ascending achievement sequence into. This parameter is set based on the total number of historical valid samples of similar general education courses. To set it, first calculate the median of the historical valid sample count, then divide this median by the minimum number of samples required to retain in each segment. The integer part of the quotient is used as the preset interval quantity parameter. If the quotient is less than 1, the parameter is set to 1. Next, the single interval carrying capacity is calculated as: total number of elements in the ascending achievement sequence / preset interval quantity parameter. If the result is an integer, the quota is calculated based on that integer quantity. The data is divided into equal segments from front to back. If the result is not an integer, the integer part is taken as the base quota, and the remaining unallocated elements are added to the first few segments from front to back, with each added segment adding one element. If the preset interval quantity parameter is greater than the total number of elements in the ascending achievement sequence, the interval quantity parameter is adjusted to be the same as the total number of elements. Then, an equal data segment set is formed according to the segmentation result, and the mean value in each segment is calculated. The segment mean value = the sum of all values ​​in the current segment / the number of elements in the current segment. Finally, the mean values ​​of all segments are written into the capability achievement discrete node in the order of segmentation, as the capability feature input in the subsequent multi-source feature stitching.

[0026] The specific steps for obtaining the difficulty compensation score sequence are as follows: S311: Extract raw score data from the general education course assessment database, perform a summation operation on all the scores in the raw score data, divide the summation value by the total number of students to obtain the course average score, calculate the difference between the raw score data and the course average score, perform a square operation on each difference to construct a set of score square differences, add all the elements in the score square difference set and divide by the total number of students to obtain the course difficulty index. Raw grade data is extracted from the general education course assessment database by course number, semester number, and class number. The raw grade data refers to the grade table formed by student identification and corresponding course score. After extraction, null, duplicate, and invalid values ​​are checked. Null grades are deleted. Records with duplicate grades and the same student identification are retained only with the final confirmed score. Grades outside the preset full score threshold range are marked as abnormal and deleted. Then, all valid grades are summed and divided by the total number of students to obtain the course average score: Course Average Score = Sum of all scores in the raw grade data / Total number of students. Finally, the difference between each student's score and the course average score is read and squared. Then, sum up all the squared differences of grades and divide by the total number of students. The course difficulty index = sum of squared differences of grades / total number of students. This course difficulty index represents the degree of dispersion of the current general education course grades around the course average grade. The larger the value, the more dispersed the grade distribution within the same course. The smaller the value, the more concentrated the grade distribution within the same course. If the total number of students is equal to 1, the course average grade is taken as this unique grade, and the course difficulty index is marked as 0.

[0027] S312: Obtain the preset benchmark difficulty index, perform a subtraction calculation based on the course difficulty index and the preset benchmark difficulty index to obtain the difficulty difference, extract the absolute value of the difficulty difference as the absolute value of the offset, determine the positive or negative attribute of the value of the difficulty difference to extract the offset direction factor, and perform a multiplication operation between the absolute value of the offset and the offset direction factor to generate the difficulty offset. The system reads a preset benchmark difficulty index, which serves as a reference value for the dispersion level of historical grades in similar general education courses. To set this index, it first extracts all course difficulty indices from historical grade tables of similar general education courses, then sorts them by value from smallest to largest, and uses the middle value as the benchmark. If the number of historical courses is even, it takes the average of the two middle values; if the number of historical courses is odd, it takes the middle value. Then, it calculates the difference between the current course difficulty index and the preset benchmark difficulty index: Difficulty Difference = Current Course Difficulty Index - Preset Benchmark Difficulty Index. First, the absolute value of the difficulty difference is read as the absolute value of the offset, and the positive or negative attribute of the difficulty difference is determined. When the difficulty difference is greater than 0, the offset direction factor is recorded as 1; when the difficulty difference is equal to 0, the offset direction factor is recorded as 0; when the difficulty difference is less than 0, the offset direction factor is recorded as -1. Then, the difficulty offset is calculated as: Difficulty offset = Absolute value of offset × Offset direction factor. When the difficulty difference is greater than 0, a positive difficulty offset is obtained; when the difficulty difference is equal to 0, a 0 difficulty offset is obtained; when the difficulty difference is less than 0, a negative difficulty offset is obtained. This result is then written into the performance compensation calculation table.

[0028] S313: Call the original score data and the course average score to calculate the difference to obtain the individual score deviation, extract the preset difficulty reverse weighting parameter, multiply the individual score deviation by the difficulty offset and the difficulty reverse weighting parameter to calculate the score correction increment, add the score correction increment to the original score data, obtain the global correction score set, reorganize and sort the global correction score set according to the personnel identity, and establish the difficulty compensation score sequence. The difficulty inverse weighting parameter is set based on the remaining score range between the preset full score threshold and the average course score, and the proportional mapping relationship constructed by the overall score variance of similar general education courses over the years. The specific setting logic is as follows: calculate the difference between the preset full score threshold and the average course score, and extract the score tolerance margin; at the same time, obtain the overall score variance of similar general education courses over the years as the distribution dispersion benchmark; divide the score tolerance margin by the distribution dispersion benchmark to obtain the quotient ratio, and use the quotient ratio as the difficulty inverse weighting parameter, so as to characterize the dynamic compression margin when approaching the full score boundary during the score compensation process, and avoid the score overflow from the effective assessment score range due to the inverse compensation calculation; The system reads the original grade data and the average course grade, and calculates the deviation of each student's grade from the average course grade. The individual grade deviation is calculated as: Current student grade - Average course grade. Next, it reads the preset full-score threshold and the average course grade, calculating the grade tolerance margin as: Preset full-score threshold - Average course grade. Then, it reads the variance of the overall grades of similar general education courses over the years as the distribution dispersion benchmark. The overall grade variance refers to the average of the squared deviations of historical course grades from their respective average course grades. Finally, the difficulty inverse weighting parameter is calculated as: Grade tolerance margin / Distribution dispersion benchmark. When the distribution dispersion benchmark is greater than 0, the above ratio is used; when the distribution dispersion benchmark is equal to 0, the difficulty inverse weighting parameter is set to 0. This parameter represents the remaining margin between the current average course grade and the full-score boundary. The proportion relative to historical dispersion is calculated. Then, for each student, the grade correction increment is calculated: Grade Correction Increment = Individual Grade Deviation × Difficulty Offset × Difficulty Inverse Weighting Parameter. The grade correction increment is then added to the original grade to obtain the corrected grade: Corrected Grade = Original Grade + Grade Correction Increment. If the corrected grade is greater than the preset full-score threshold, the grade is adjusted to the preset full-score threshold. If the corrected grade equals the preset full-score threshold, it remains unchanged. If the corrected grade is between 0 and the preset full-score threshold, it is directly retained. If the corrected grade equals 0, it remains unchanged. If the corrected grade is less than 0, it is corrected to 0. Finally, all corrected grades are reordered according to student identity to form a difficulty-compensated grade sequence, maintaining the same index order as the preceding ability achievement data.

[0029] The specific steps for obtaining the evaluation period parameters are as follows: S411: Collect continuous scoring data sequences through the general education evaluation portal, perform subtraction operations on adjacent node values ​​within the continuous scoring data sequence to construct a scoring change sequence, extract all differences within the scoring change sequence and sum them to obtain the total difference, divide the total difference by the total number of elements in the scoring change sequence to calculate the mean, and establish a scoring change reference benchmark. Continuous rating data sequences are collected through the general education evaluation portal. These sequences are timestamp-ordered evaluation score records. The data is collected by exporting student rating data tables from the evaluation webpage or mobile backend and sorting them by submission time from earliest to latest. Next, adjacent rating nodes are read sequentially, and the difference between the later and earlier rating values ​​is calculated: rating change = current rating value - previous rating value. All adjacent rating changes are written into the rating change sequence in chronological order. Then, all differences in the rating change sequence are summed and divided by the total number of elements in the sequence. The rating change reference base = sum of all differences in the rating change sequence / total number of elements in the rating change sequence. If the total number of elements in the rating change sequence is greater than 0, the above calculation is performed; if the total number of elements is 0, the rating change reference base is recorded as 0. If multiple rating records correspond to the same timestamp, they are first refined and sorted according to submission order before the extraction of differences between adjacent nodes is performed.

[0030] S412: Obtain the preset judgment ratio factor, multiply the score change reference benchmark with the preset judgment ratio factor to calculate the fluctuation judgment threshold, extract the segments in the score change sequence that are continuously lower than the fluctuation judgment threshold to form continuous stable segments, read the start and end timestamps of the continuous stable segments and subtract them to obtain the duration span, retain the continuous stable segments associated with the judgment benchmark whose duration span is greater than the preset length, and establish an effective period candidate set. The determination ratio factor is set based on the background fluctuation characteristics of the historical teaching evaluation dataset within a known steady-state teaching period. By extracting multiple sets of historical steady-state teaching evaluation sample sequences, the ratio of the local score dispersion to the change in the global score mean within each set sequence is calculated to construct a set of background error ratio distributions. Based on a statistical probability density model, the convergence upper bound critical coefficient value of the set of background error ratio distributions under a pre-set reliability condition is extracted. The critical coefficient value is used as the maximum tolerance scale limit for the fluctuation of the real evaluation after excluding the interference of normalized random scoring, and it is established as the determination ratio factor. The system reads the preset judgment ratio factor and the benchmark for score changes. The fluctuation judgment threshold is calculated as: benchmark for score changes × preset judgment ratio factor. The preset judgment ratio factor refers to the maximum allowable amplification factor for score changes within a historical steady-state teaching cycle. It is set by first extracting multiple sets of score sequences from the historical evaluation dataset within a known steady-state teaching cycle, then calculating the dispersion of adjacent score differences within each sequence and the overall mean change of that sequence. Local score dispersion = sum of absolute values ​​of score changes in the current group / number of score changes in the current group; global mean change = |mean score at the end of the current group - mean score at the beginning of the current group|; baseline error ratio = local score dispersion / global mean change. When the global mean change is 0, the baseline error ratio of that group is recorded as the maximum recorded value corresponding to the local score dispersion of the current group. Subsequently, all baseline error ratios are sorted from smallest to largest, and the upper bound value of the corresponding position is read according to the preset reliability as the judgment ratio factor. The preset reliability is determined by the steady-state cycle recorded in the historical evaluation data. The lower limit of the proportion of all valid records is determined. Then, each difference in the score change sequence is compared with the fluctuation judgment threshold. If the current difference is less than the fluctuation judgment threshold, and the previous difference is also less than or equal to the fluctuation judgment threshold, it is merged into the current continuous stable segment. If the previous difference is greater than the fluctuation judgment threshold, the position of the current difference is used as the starting point of the new continuous stable segment. If the current difference is equal to the fluctuation judgment threshold, the position of the current difference is merged into the continuous stable segment. If the current difference is greater than the fluctuation judgment threshold, the current continuous stable segment ends and waits for the next starting point that meets the conditions. Then, the start timestamp and end timestamp of each continuous stable segment are read, and the duration span is calculated. The duration span = the end time of the continuous stable segment - the start time of the continuous stable segment. The duration span is then compared with the preset length judgment benchmark. If the duration span is greater than the preset length judgment benchmark, the segment is retained. If the duration span is equal to the preset length judgment benchmark, the segment is retained. If the duration span is less than the preset length judgment benchmark, the segment is deleted. Finally, a candidate set of valid periods is formed.

[0031] S413: For the effective period candidate set, extract the highest score and the lowest score in each period and perform subtraction to obtain the interval fluctuation range. Divide the interval fluctuation range by the corresponding duration span to calculate the fluctuation ratio coefficient. Sort all the calculated fluctuation ratio coefficients in ascending order and extract the first value of the queue as the minimum value item. Read the period length data mapped by the minimum value item to generate the evaluation period parameters. Each period in the valid period candidate set is read item by item, and all scores within that period are read. The highest and lowest scores are identified, and the interval fluctuation range is calculated as: highest score within the period - lowest score within the period. Then, the duration span corresponding to that period is read, and the fluctuation ratio coefficient is calculated as: fluctuation ratio coefficient = interval fluctuation range / duration span. When the duration span is greater than 0, the calculation is performed as described above; when the duration span is equal to 0, the fluctuation ratio coefficient for that period is recorded as the maximum recorded value corresponding to the interval fluctuation range itself. Next, all fluctuation ratio coefficients are sorted from smallest to largest, and the first value after sorting is read as the minimum value. If there is only one minimum value, the length of the period corresponding to that minimum value is directly read as the evaluation period parameter. If there are multiple identical minimum values, the period lengths corresponding to these minimum values ​​are compared. If the period length is shorter, the shorter length is directly used as the evaluation period parameter; if the period lengths are equal, their common length is used as the evaluation period parameter. The evaluation period parameter refers to the time length corresponding to each segment when subsequent continuous scoring data is segmented.

[0032] The specific steps for obtaining the results of the general education quality assessment are as follows: S511: Perform equidistant segmentation on the continuous scoring data sequence according to the evaluation cycle parameter to construct a set of scoring sub-segments. Extract the scoring values ​​covered by each sub-segment of the scoring sub-segment set and perform cumulative operation to obtain the sub-segment sum. Divide the sub-segment sum by the number of elements in the corresponding sub-segment to calculate the average value and generate the segmented scoring mean. The evaluation period parameter is read, and the continuous scoring data sequence is segmented sequentially from the start time according to the time length corresponding to the parameter, forming multiple sets of scoring segments with the same time span. Then, for each scoring segment, all scores within that time period are read and the sum is calculated. This sum is then divided by the number of scores in the current scoring segment. The segment average score = sum of scores in the current scoring segment / number of scores in the current scoring segment. If the number of scores in the current scoring segment is greater than 0, the segment average score is calculated directly. If the number of scores in the current scoring segment is equal to 0, the segment is marked as empty, and the average scores of the preceding and following non-empty segments are read. If non-empty segments exist on both sides, their averages are written into the current empty segment. If only the preceding segment has a non-empty segment, the preceding average score is written. If only the following segment has a non-empty segment, the following average score is written. If neither the preceding nor following segment has a non-empty segment, 0 is written. Finally, the segment average scores corresponding to all scoring segments are written into the segment average score sequence in chronological order.

[0033] S512: Perform column-wise concatenation of discrete nodes of ability achievement, difficulty compensation score sequence and segment score mean according to preset index to extract multi-source feature matrix. Divide each element in the multi-source feature matrix by preset extreme value constant to perform normalization mapping, compress numerical values ​​and extract multi-dimensional evaluation feature vector. The algorithm reads discrete nodes representing ability achievement, sequence of difficulty compensation scores, and average segmented scores, aligns them according to a preset index, and then performs column-wise concatenation. The preset index refers to using a consistent sequential identifier for the three types of data within the same course number and statistical batch. Subsequently, discrete nodes representing ability achievement, difficulty compensation scores, and average segmented scores at the same index position are written to the same row, forming a multi-source feature matrix. Then, a preset extreme value constant is read, and the ratio of each element in the matrix to the preset extreme value constant is calculated. The normalized eigenvalue is calculated as: current matrix element / preset extreme value constant. A positive ratio is written when the current matrix element is greater than 0, 0 is written when the current matrix element is equal to 0, and a ratio is written when the current matrix element is less than 0. The negative value ratio is written in time; the preset extreme value constant refers to the reference upper bound used when scaling the three types of features uniformly. When setting it, first read all the absolute values ​​of the three types of features in the historical general education courses, then find the maximum absolute value, and add a 10% margin on the basis of the maximum absolute value. The preset extreme value constant = the historical maximum absolute value × 1.1; if the absolute value of an element in the current matrix is ​​greater than the preset extreme value constant, then update the preset extreme value constant to 1.1 times the current maximum absolute value, and then re-execute the normalization of all elements in the current batch; finally, combine the three normalized values ​​of each row into a multi-dimensional evaluation feature vector in a fixed order, and write it into the feature vector table in the order of course number and semester number.

[0034] S513: Call the preset support vector regression model, import the multidimensional evaluation feature vector into the support vector regression model, call the kernel function associated with the support vector regression model to calculate the inner product value of the multidimensional evaluation feature vector and the support vector, multiply the inner product value by the corresponding weight coefficient and add the bias constant to perform regression estimation operation, extract the estimated value to generate the general education quality assessment result. The process of performing regression estimation by multiplying the inner product value by the corresponding weight coefficient and adding the bias constant is as follows: The kernel function associated with the support vector regression model is determined to be a Gaussian function, and the support vector regression model is set by minimizing the error of the historical evaluation vector set; the squared distance value is obtained by calculating the square of the absolute value of the difference between the multidimensional evaluation feature vector and the support vector; the kernel function parameters are obtained, and the squared distance value is multiplied by the negative of the kernel function parameters to obtain the decay value. The inner product value is obtained by exponentiation of the decay value over the natural base, and the kernel function parameters are set according to the reciprocal of the support vector distribution distance; the corresponding weight coefficients and bias constants are read, with the weight coefficients set according to the difference of the penalty factor in the optimization solution, and the bias constants set according to the average difference between the actual output of the support vectors and the expected inner product value; the inner product value is multiplied by the corresponding weight coefficient and then summed to obtain the comprehensive mapping value; the comprehensive mapping value is added to the bias constant to obtain the estimated value. The pre-built support vector regression model is read. Support vector regression is a data processing structure that uses the numerical relationship between historical multidimensional evaluation feature vectors and their corresponding teaching quality results to complete regression estimation. This model is executed in Python. First, it reads the multidimensional evaluation feature vectors and their corresponding actual teaching quality results from historical evaluation records. Then, it divides the training set and validation set into an 8:2 ratio. Subsequently, it reads each feature vector and its corresponding result value from the training set, and sets the initial kernel parameters, weight coefficients, bias constants, and penalty factors. The kernel parameters control the magnitude of distance decay between vectors. When setting these parameters, the squared distance between any two feature vectors in the training set is calculated first. The squared distance is calculated as follows: [Equation missing - likely a formula for calculating the distance between two feature vectors]. The kernel parameter is calculated by summing the squared differences of the dimensions and then averaging the squared distances of all distances. The kernel parameter is set to 1 / average squared distance. If the average squared distance is greater than 0, it is calculated as described above; if the average squared distance is equal to 0, the kernel parameter is set to 1. The weight coefficients represent the contribution of each support vector. The bias constant represents the overall shift between the comprehensive mapping result and the actual output. The penalty factor is the control value for parameter correction when the error exceeds the allowable range. The initial value of the penalty factor is the reciprocal of the standard deviation of the historical evaluation results. If the standard deviation of the historical evaluation results is greater than 0, it is calculated using the reciprocal; if it is equal to 0, it is set to 1. Subsequently, the squared distances between the feature vectors and support vectors in the training set are read round by round, and the kernel mapping values ​​are obtained in Gaussian form. Then, multiply each kernel mapping value by its corresponding weight coefficient, sum them up, and add the bias constant to obtain the estimated value. Next, compare the estimated value with the actual teaching quality result value. The error value = |estimated value - actual teaching quality result value|. If the cumulative error value of the current round is less than the cumulative error value of the previous round, retain the weight coefficient, bias constant, and penalty factor of the current round. If the cumulative error value of the current round is equal to the cumulative error value of the previous round, keep the current parameters unchanged and continue to the next round. If the cumulative error value of the current round is greater than the cumulative error value of the previous round, reduce the adjustment range of the weight coefficient by a fixed step size and simultaneously reduce the penalty factor, then recalculate. When the cumulative error value no longer decreases after several consecutive rounds, stop adjusting and fix the parameters. Subsequently, generate the current batch... The multidimensional evaluation feature vectors are input one by one into the support vector regression model with fixed parameters. The estimated values ​​are calculated in the order of the aforementioned squared distance, kernel mapping value, sum of products, and bias stacking, and the estimated values ​​are written into the general education quality evaluation result table. Finally, the validation set feature vectors are compared with the actual result values. The validation error = |validation estimated value - validation actual value|. If the validation error is greater than the preset tolerance range, the kernel parameters, weight coefficients, and penalty factors are readjusted. If the validation error is equal to the preset tolerance range, the current parameters are fixed. If the validation error is less than the preset tolerance range, the current parameters are fixed and the general education quality evaluation result is output.

[0035] A teaching quality assessment system based on general education, used to implement the aforementioned teaching quality assessment method based on general education, includes: The benchmark screening module extracts the sequence of general education ability achievement, calculates the median value of the ability and the difference between each item and the median value of the ability, constructs the median difference sequence, sorts and selects the set of values ​​within the preset ratio range on both sides as the stable candidate set, and calculates the mean and standard deviation to generate the mean and standard deviation range of the stable interval. The node integration module performs outlier detection on the capability achievement sequence based on the mean of the stable interval and the standard offset range, removes values ​​that exceed the outlier detection threshold, performs equal interval division on the remaining achievement data and calculates the interval mean, and constructs discrete nodes of capability achievement. The grade correction module extracts raw grade data from the general education course assessment database and combines it with the course average grade, course difficulty index, preset benchmark difficulty and difficulty offset to generate a difficulty compensation grade sequence. The period identification module collects continuous scoring data sequences through the general education evaluation portal, constructs scoring change sequences, filters continuous stable segments and selects a set of valid period candidates, and generates evaluation period parameters. The quality assessment module calculates the average score for each segment based on the evaluation period parameters. It then inputs the discrete nodes of ability achievement, the sequence of difficulty compensation scores, and the average score for each segment into the support vector regression model to perform regression estimation and generate the general education quality assessment results.

[0036] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A teaching quality assessment method based on general education, characterized in that, Includes the following steps: S1: Extract the sequence of general education ability achievement, calculate the median value of each ability and the difference between each ability and the median value, construct the median difference sequence, sort and select the set of values ​​within the preset ratio range on both sides as the stable candidate set, and calculate the mean and standard deviation to generate the stable interval mean and standard deviation range. S2: Based on the mean of the stable interval and the standard offset range, outlier judgment is performed on the capability achievement sequence, values ​​exceeding the outlier judgment threshold are removed, and the remaining achievement data are divided into equal intervals and the interval mean is calculated to construct discrete nodes of capability achievement. S3: Extract raw grade data from the general education course assessment database, and combine it with the average course grade, course difficulty index, preset benchmark difficulty and difficulty offset to generate a difficulty-compensated grade sequence. S4: Collect continuous scoring data sequences through the general education evaluation portal, construct scoring change sequences, screen continuous stable segments and select effective period candidate sets, and generate evaluation period parameters; S5: Calculate the average score of each segment based on the evaluation period parameters, and input the discrete nodes of the ability achievement, the difficulty compensation score sequence and the average score of each segment into the support vector regression model to perform regression estimation operation and generate the general education quality assessment results.

2. The teaching quality assessment method based on general education according to claim 1, characterized in that, The specific steps for obtaining the standard offset range are as follows: S111: Obtain general education records, extract the general ability achievement sequence, sort each element in the general ability achievement sequence in ascending order of value, extract the center value corresponding to the sorting node in the queue, obtain the median value of general ability, calculate the difference between each element in the general ability achievement sequence and the median value of general ability, and establish a median value difference sequence. S112: Perform ascending sorting on the median difference sequence to construct an ascending difference sequence, obtain a preset bilateral boundary ratio, calculate the single-sided truncation number based on the preset bilateral boundary ratio, truncate and remove extreme values ​​on the ascending difference sequence based on the single-sided truncation number, extract the remaining middle position values, and construct a stable candidate set. S113: Call each value in the stable candidate set to perform a summation operation, divide the summation result by the total number of elements covered by the stable candidate set to obtain the mean of the stable interval, calculate the sum of squares of the difference between each value in the stable candidate set and the mean of the stable interval, divide the sum of squares by the total number of elements covered by the stable candidate set to perform a square root operation, and calculate the standard offset range.

3. The teaching quality assessment method based on general education according to claim 1, characterized in that, The specific steps for obtaining the discrete nodes of the capability achievement degree are as follows: S211: Call the capability achievement sequence and the mean of the stable interval, perform loop traversal extraction for each element of the capability achievement sequence, calculate the absolute value of the difference between each element and the mean of the stable interval, perform numerical aggregation processing on all extracted absolute values, and construct the achievement deviation set. S212: Based on the achievement deviation set, extract the value of each deviation, obtain the preset offset multiple and the standard offset range, calculate the corresponding product value of the preset offset multiple and the standard offset range to construct the outlier determination threshold, compare each deviation value with the outlier determination threshold, locate the associated value of the capability achievement sequence when the deviation value is greater than the outlier determination threshold, remove the associated value from the capability achievement sequence, and merge the retained items to generate the remaining achievement data; S213: Rearrange the remaining achievement data in ascending order to obtain an ascending achievement sequence, obtain a preset interval quantity parameter, divide the total number of elements in the ascending achievement sequence by the preset interval quantity parameter to calculate the single interval carrying quota, perform equal-quantity cutting on the ascending achievement sequence according to the single interval carrying quota, construct an equal-quantity data segment set, calculate the mean value of each segment in the equal-quantity data segment set, and summarize to establish discrete nodes of capability achievement.

4. The teaching quality assessment method based on general education according to claim 3, characterized in that, The offset factor is set based on obtaining a historical ability achievement sample set of the same type of general education courses, calculating the cumulative probability density distribution data of the historical ability achievement sample set under multiple standard deviation spans, combining the preset effective sample retention ratio, searching for distribution nodes in the cumulative probability density distribution data that match the effective sample retention ratio, and extracting the standard deviation multiplier corresponding to the distribution node as the offset factor.

5. The teaching quality assessment method based on general education according to claim 1, characterized in that, The specific steps for obtaining the difficulty-compensated score sequence are as follows: S311: Extract raw score data from the general education course assessment database, perform a summation operation on all the scores in the raw score data, divide the summation value by the total number of students to obtain the course average score, calculate the difference between the raw score data and the course average score, perform a square operation on each difference to construct a set of score square differences, add all the elements in the score square difference set and divide by the total number of students to obtain the course difficulty index. S312: Obtain a preset benchmark difficulty index, perform a subtraction calculation based on the course difficulty index and the preset benchmark difficulty index to obtain the difficulty difference, extract the absolute value of the difficulty difference as the absolute value of the offset, determine the positive or negative attribute of the value of the difficulty difference to extract the offset direction factor, and perform a multiplication operation between the absolute value of the offset and the offset direction factor to generate the difficulty offset. S313: The original score data and the average course score are used to calculate the difference to obtain the individual score deviation. The preset difficulty reverse weighting parameter is extracted. The individual score deviation is multiplied by the difficulty offset and the difficulty reverse weighting parameter to calculate the score correction increment. The score correction increment is superimposed on the original score data to obtain the global correction score set. The global correction score set is reorganized and sorted according to the personnel identity to establish the difficulty compensation score sequence.

6. The teaching quality assessment method based on general education according to claim 1, characterized in that, The specific steps for obtaining the evaluation period parameters are as follows: S411: Collect continuous scoring data sequences through the general education evaluation portal, perform subtraction operations on adjacent node values ​​within the continuous scoring data sequence to construct a scoring change sequence, extract all differences within the scoring change sequence and sum them to obtain the total difference, divide the total difference by the total number of elements in the scoring change sequence to calculate the mean, and establish a scoring change reference benchmark. S412: Obtain a preset judgment ratio factor, multiply the score change reference benchmark by the preset judgment ratio factor to calculate the fluctuation judgment threshold, extract the segments in the score change sequence that are continuously lower than the fluctuation judgment threshold to form a continuous stable segment, read the start and end timestamps of the continuous stable segment and subtract them to obtain the duration span, retain the continuous stable segments associated with the duration span greater than the preset length judgment benchmark, and establish an effective period candidate set. S413: For the effective period candidate set, extract the highest score and the lowest score in each period and perform subtraction to obtain the interval fluctuation range. Divide the interval fluctuation range by the corresponding duration span to calculate the fluctuation ratio coefficient. Sort all the calculated fluctuation ratio coefficients in ascending order and extract the first value of the queue as the minimum value item. Read the period length data mapped by the minimum value item to generate the evaluation period parameter.

7. The teaching quality assessment method based on general education according to claim 6, characterized in that, The determination ratio factor is set based on the background fluctuation characteristics of the historical evaluation dataset within a known steady-state teaching period. By extracting multiple sets of historical steady-state evaluation sample sequences, the ratio of the local score dispersion to the global score mean change within each set is calculated to construct a set of background error ratio distributions. Based on a statistical probability density model, the convergence upper bound critical coefficient value of the set of background error ratio distributions under a pre-set reliability condition is extracted. The critical coefficient value is used as the maximum tolerance scale limit for the fluctuation of the real evaluation after excluding the interference of normalized random scoring, and it is established as the determination ratio factor.

8. The teaching quality assessment method based on general education according to claim 1, characterized in that, The specific steps for obtaining the general education quality assessment results are as follows: S511: Perform equidistant segmentation on the continuous scoring data sequence according to the evaluation cycle parameters to construct a set of scoring sub-segments. Extract the scoring values ​​covered by each sub-segment of the scoring sub-segment set and perform an accumulation operation to obtain the sub-segment sum. Divide the sub-segment sum by the number of elements in the corresponding sub-segment to calculate the average value and generate the segmented scoring mean. S512: The discrete nodes of the ability achievement degree, the difficulty compensation score sequence and the average of the segmented scores are column-wise concatenated according to the preset index to extract a multi-source feature matrix. Each element in the multi-source feature matrix is ​​divided by a preset extreme value constant to perform normalization mapping, and the numerical values ​​are compressed to extract a multi-dimensional evaluation feature vector. S513: Call the preset support vector regression model, import the multidimensional evaluation feature vector into the support vector regression model, call the kernel function associated with the support vector regression model to calculate the inner product value of the multidimensional evaluation feature vector and the support vector, multiply the inner product value by the corresponding weight coefficient and add the bias constant to perform regression estimation operation, extract the estimated value to generate the general education quality assessment result.

9. The teaching quality assessment method based on general education according to claim 8, characterized in that, The process of performing regression estimation by multiplying the inner product value by the corresponding weight coefficient and adding the bias constant specifically involves: determining the kernel function associated with the support vector regression model as a Gaussian function; setting the support vector regression model by minimizing the error of the historical evaluation vector set; calculating the square of the absolute value of the difference between the multidimensional evaluation feature vector and the support vector to obtain the squared distance value; obtaining the kernel function parameter; multiplying the squared distance value by the negative of the kernel function parameter to obtain the decay value; raising the natural base to the power of the decay value to obtain the inner product value; setting the kernel function parameter based on the reciprocal of the support vector distribution distance; reading the corresponding weight coefficient and bias constant; setting the corresponding weight coefficient according to the difference of the penalty factor in the optimization solution; setting the bias constant according to the average difference between the actual output of the support vector and the expected inner product value; multiplying the inner product value by the corresponding weight coefficient and summing the results to obtain the comprehensive mapping value; and performing an addition operation on the comprehensive mapping value and the bias constant to obtain the estimated value.

10. A teaching quality assessment system based on general education, characterized in that, The system is used to implement the teaching quality assessment method based on general education as described in any one of claims 1-9, and the system comprises: The benchmark screening module extracts the sequence of general education ability achievement, calculates the median value of the ability and the difference between each item and the median value of the ability, constructs the median difference sequence, sorts and selects the set of values ​​within the preset ratio range on both sides as the stable candidate set, and calculates the mean and standard deviation to generate the mean and standard deviation range of the stable interval. The node integration module performs outlier detection on the capability achievement sequence based on the mean of the stable interval and the standard offset range, removes values ​​that exceed the outlier detection threshold, performs equal interval division on the remaining achievement data and calculates the interval mean, and constructs discrete nodes of capability achievement. The grade correction module extracts raw grade data from the general education course assessment database and combines it with the course average grade, course difficulty index, preset benchmark difficulty and difficulty offset to generate a difficulty compensation grade sequence. The period identification module collects continuous scoring data sequences through the general education evaluation portal, constructs scoring change sequences, filters continuous stable segments and selects a set of valid period candidates, and generates evaluation period parameters. The quality assessment module calculates the average score for each segment based on the evaluation period parameters, and inputs the discrete nodes of the ability achievement, the difficulty compensation score sequence, and the average score for each segment into the support vector regression model to perform regression estimation and generate the general education quality assessment results.

Images