AI-based teaching quality control and adjustment method

Abstract

The invention discloses a teaching quality control and adjustment method based on AI, and relates to the technical field of AI. The method comprises the steps that an AI quality analysis terminal analyzes a teaching cleaning optimization data set to obtain a teaching effect influence coefficient, a student participation coefficient and a teaching resource adaptation coefficient; inputting the teaching effect influence coefficient, the student participation coefficient and the teaching resource adaptation coefficient into a preset teaching quality comprehensive evaluation model, and outputting a teaching quality comprehensive evaluation index; and the teaching adjustment strategy generation terminal generates a corresponding teaching adjustment strategy according to the teaching quality comprehensive index. According to the invention, by setting three levels of teaching adjustment strategies, different strategies are triggered according to the comprehensive indexes of the teaching quality, targeted strengthening, weakening or reconstruction is carried out on the teaching effect, the student participation degree and the teaching resource adaptation, the refined hierarchical regulation and control of the teaching quality are realized, and the efficiency and the pertinence of teaching regulation and control are improved.

Description

Technical Field

[0001] The present invention relates to the field of AI technology, and in particular to an AI-based teaching quality control and adjustment method. Background Art

[0002] With the rapid adoption of online education platforms and the widespread use of intelligent teaching tools, teaching environments are gradually shifting from traditional classrooms to digital and networked ones. Vast amounts of teaching data are continuously generated through online interactions, covering multiple dimensions such as student behavior, resource usage, and classroom interaction. However, traditional teaching quality control methods, which primarily rely on manual spot checks, periodic evaluations, and static indicator analysis, struggle to capture teaching dynamics in real time. This leads to delayed evaluations, limited coverage, and an inability to adapt to the rapidly growing demand for personalized teaching.

[0003] Currently, existing technologies mostly focus on single-dimensional data analysis, such as predicting learning outcomes based on student performance or assessing the usage of teaching materials based on resource access volume. While some systems can achieve basic data collection and report generation, they lack significant deficiencies in multi-source data integration, real-time feedback, and dynamic strategy adjustment. Furthermore, most methods lack a collaborative analysis of teaching effectiveness, student engagement, and resource compatibility, resulting in one-sided strategy recommendations and difficulty in systematically improving teaching quality.

[0004] The core problems facing the current teaching quality control system are: first, data collection is scattered across multiple systems, failing to integrate full-link information on student behavior and teaching resources, affecting the comprehensiveness of the evaluation; second, the quality assessment model relies on manually set weights, is highly subjective and cannot be adaptively optimized, resulting in biased evaluation results; third, the adjustment strategy is rigid and cannot be calibrated in real time according to dynamic changes in teaching quality, making it difficult to achieve stratified and precise intervention.

[0005] Therefore, it is urgent to invent an AI-based teaching quality control and adjustment method to solve the above problems. Summary of the Invention

[0006] The purpose of the present invention is to provide an AI-based teaching quality control and adjustment method to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an AI-based teaching quality control and adjustment method, comprising a teaching data acquisition terminal, a teaching data preprocessing terminal, an AI quality analysis terminal, a teaching adjustment strategy generation terminal, and a feedback execution terminal, specifically comprising the following steps: S1. The teaching data acquisition terminal collects data from each interactive node of the online teaching platform to obtain an original teaching data set, which includes a student data set and a teaching resource data set; S2, the teaching data preprocessing terminal processes the original teaching data set to obtain a teaching cleansing and optimization data set; S3. The AI ​​quality analysis terminal analyzes the teaching cleaning optimization data set to obtain a teaching effect impact coefficient, a student participation coefficient, and a teaching resource adaptation coefficient, and inputs the teaching effect impact coefficient, the student participation coefficient, and the teaching resource adaptation coefficient into a preset teaching quality comprehensive evaluation model to output a teaching quality comprehensive evaluation index; S4. The teaching adjustment strategy generating terminal generates a corresponding teaching adjustment strategy according to the teaching quality comprehensive index; S5. The feedback execution terminal feeds back the generated teaching adjustment strategy to the relevant teaching subjects, and regularly re-evaluates the teaching quality and dynamically adjusts the strategy according to the evaluation results.

[0008] Preferably, the student data set includes homework scores, homework completion time, test scores, test length, classroom interaction effectiveness, classroom interaction duration, number of students involved in the interaction, total number of homework assignments, total number of tests, total number of classroom interactions, online learning duration, and login frequency; the teaching resource data set includes resource access depth, resource relevance, resource release time, number of teaching resources accessed, quality rating of teaching resources, update time interval of teaching resources, proportion of students applicable to teaching resources, difficulty coefficient of teaching resources, total number of teaching resources, and total number of effective teaching resources.

[0009] Preferably, the teaching effect influence coefficient is specifically: , Among them, S i To represent the score of the i-th assignment, W i is the weight of the i-th job, t i is the completion time of the i-th job, T max is the longest job completion time in the statistical period, C j is the result of the jth test, T j is the length of time for the jth test, △C j is the difference between the jth test score and the previous one, C avg is the average value of the change in grades during the statistical period, I k is the effectiveness of the k-th classroom interaction, L k is the duration of the k-th classroom interaction, N k is the number of students involved in the kth interaction, n is the total number of homework, m is the total number of tests, p is the total number of classroom interactions, is the average score of the assignments, is the average of the test scores, is the average value of classroom interaction effectiveness.

[0010] Preferably, the student participation coefficient is specifically: , in, , , , , where S1 is the online learning behavior item, S2 is the resource access quality item, D1 is the scale correction item for learning time, and D2 is the fluctuation correction item for resource depth. l is the online learning time on day 1, F l is the login frequency on the first day, A r is the rth resource access depth, D r is the relevance of the rth resource, d r is the release time of the rth resource, D max is the longest release time of resources in the statistical period, q is the number of statistical days, s is the number of teaching resources visited, is the average daily online learning time, is the average teaching resource access depth, σ A is the standard deviation of the access depth of teaching resources, e is the Euler number, and ln is the logarithm with base e.

[0011] Preferably, the teaching resource adaptation coefficient is specifically: , Among them, Q t is the quality score of the t-th teaching resource, K t is the update time interval of the t-th teaching resource, H v is the proportion of students applicable to the vth teaching resource, M v is the difficulty coefficient of the vth teaching resource, u is the total number of teaching resources, w is the total number of effective teaching resources, is the average value of the teaching resource quality rating, is the average of the applicable student proportions.

[0012] Preferably, the comprehensive teaching quality evaluation model is specifically: , Among them, E is the teaching effect influence coefficient, P is the student participation coefficient, R is the teaching resource adaptation coefficient, δ is the bias term, α, β and γ are weight coefficients, α+β+γ=1 and α, β and γ∈[0,1].

[0013] Preferably, the teaching adjustment strategy specifically includes: When the comprehensive teaching quality index QI ≥ m1, the first-level strategy is triggered; When the comprehensive teaching quality index m2 ≤ QI < m1, the secondary strategy is triggered; When the comprehensive teaching quality index QI < m2, the tertiary strategy is triggered.

[0014] Preferably, the primary strategy is to strengthen the teaching effect through experience replication and promotion, high-order ability cultivation, and dynamic monitoring threshold: improve student participation through incentive mechanism upgrade and precise resource push; adapt teaching resources through benchmark resource precipitation and forward-looking update; The secondary strategy is to supplement and strengthen the teaching effect through homework optimization, test adjustment, and interaction enhancement; improve student participation through time management intervention and resource stickiness optimization; calibrate the adaptation of teaching resources through quality and timeliness optimization and difficulty and universality adjustment; The tertiary strategy is to reconstruct the teaching effect through the transformation of the teaching mode; activate student participation through personalized incentive programs and social learning intervention; reconstruct the adaptation of teaching resources through comprehensive resource evaluation and replacement and intelligent resource push.

[0015] Preferably, the process of dynamic adjustment is as follows: The feedback execution terminal starts the teaching quality review mechanism at a preset cycle, re-collects teaching data, and obtains a new comprehensive teaching quality evaluation index QI through the steps of S2 and S3 new , if QI new and the absolute value of the difference from the historical comprehensive teaching quality evaluation index QI old exceeds the preset threshold, the strategy adjustment logic is triggered.

[0016] Preferably, the strategy adjustment logic includes dynamic threshold calibration, specifically: , , where QI′ is the average value of the historical comprehensive teaching quality evaluation index, and σ QI is the standard deviation of the historical comprehensive teaching quality evaluation index.

[0017] The technical effects and advantages of the present invention: The present invention collects teaching raw data including student data sets and teaching resource data sets from each interaction node of the online teaching platform through a teaching data collection terminal, and obtains the teaching effect influence coefficient, student participation coefficient, and teaching resource adaptation coefficient through preprocessing and AI analysis, and then inputs them into a comprehensive evaluation model to output a comprehensive index, realizing multi-dimensional and precise evaluation of teaching quality, providing a scientific basis for teaching adjustment; The present invention sets a three-level teaching adjustment strategy, triggering different strategies according to the comprehensive teaching quality index, and specifically strengthening, improving or reconstructing teaching effects, student participation and teaching resource adaptation, thereby achieving refined hierarchical regulation of teaching quality and improving the efficiency and pertinence of teaching regulation. The present invention regularly re-evaluates the teaching quality through the feedback execution terminal, dynamically adjusts the strategy according to the evaluation results, and combines the strategy adjustment logic of dynamic calibration of the threshold value to enable the teaching adjustment strategy to be optimized in real time according to the actual teaching situation, thereby enhancing the system's adaptability and accuracy to changes in teaching quality and ensuring the effectiveness and sustainability of teaching quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a schematic diagram of the device connection of the present invention.

[0019] Figure 2 The present invention is a flowchart of the steps for implementing the method. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] The present invention provides Figure 1 An AI-based teaching quality control and adjustment method is shown, including a teaching data acquisition terminal, a teaching data preprocessing terminal, an AI quality analysis terminal, a teaching adjustment strategy generation terminal and a feedback execution terminal.

[0022] It should be noted that the teaching data acquisition terminal supports multi-threaded concurrent capture of online platform interaction data through high I / O performance configuration, and is paired with raw data storage devices such as MySQL or MongoDB database servers and SSD hard drives to achieve real-time storage of unprocessed data; The teaching data preprocessing terminal relies on the data processing server to run the cleaning algorithm and feature engineering logic. The mid-to-high-end CPU configuration supports parallel computing to process massive data. When the data volume is large, it can be expanded to Hadoop or Spark distributed computing clusters to improve processing efficiency. Intermediate data storage devices, such as HDFS or Hive data warehouses, are responsible for temporarily storing the cleaned and optimized data sets, building a processing hub from raw data cleaning to standardized output. The AI ​​quality analysis terminal uses AI computing servers as its computing power core, GPU servers such as NVIDIA A100 / H100, CPU servers adapted to traditional machine learning algorithms, high-speed NVMe SSDs to store training data and model parameters, and computing power scheduling equipment such as Kubernetes to achieve dynamic allocation of cluster resources. It uses a computing engine and a comprehensive evaluation model to complete intelligent analysis of the three major coefficients and teaching quality index. The teaching adjustment strategy generation terminal is configured with an ordinary server or virtual machine, such as a cloud server EC2, which supports high-concurrency strategy generation requests and is equipped with a three-level strategy template for storage, such as SQL Server storage strategy rules and JSON files for storing specific measures. It is connected to the AI ​​analysis terminal to obtain QI values ​​in real time and trigger strategy generation. The feedback execution terminal is equipped with a message queue service, such as RabbitMQ or Kafka, to support high-throughput message sending. The scheduled task server starts a dynamic review process, and the execution record database, such as Elasticsearch, stores policy logs, together forming a closed-loop execution system for policy implementation, effect tracking and dynamic adjustment.

[0023] The present invention provides Figure 2 The AI-based teaching quality control and adjustment method shown includes the following steps: S1. The teaching data acquisition terminal collects data from each interactive node of the online teaching platform to obtain an original teaching data set, which includes a student data set and a teaching resource data set; Furthermore, in the above technical solution, the student data set includes homework scores, homework completion time, test scores, test duration, effectiveness of classroom interaction, duration of classroom interaction, number of students involved in interaction, total number of homework, total number of tests, total number of classroom interactions, online learning duration and login frequency; the teaching resource data set includes resource access depth, resource relevance, resource release time, number of accessed teaching resources, quality score of teaching resources, update time interval of teaching resources, proportion of students applicable to teaching resources, difficulty coefficient of teaching resources, total number of teaching resources and total number of effective teaching resources.

[0024] It should be noted that the way to collect the score of the homework is that after the student submits the homework, the system automatically corrects and records the score, such as the automatic scoring function of the online answering platform; the way to collect the completion time of the homework is the time interval from the student starting to answer the questions to submitting the homework, which is timed and stored in real time by the system background; the way to collect the results of the test is that after the test is over, the system automatically generates the results according to the answer situation, supports automatic scoring of objective questions, and manual or AI-assisted scoring of subjective questions; the way to collect the time length of the test is the time length from the student starting the test to submitting the test paper, which is recorded by the system through the timing function; the way to collect the effectiveness of the classroom interaction is to analyze the classroom chat records through the natural language processing (NLP) algorithm Recording, voice-to-text content, identifying effective interactions related to teaching objectives, eliminating irrelevant speeches, and calculating the effectiveness according to the proportion of effective content; the method for collecting the duration of the classroom interaction is to count the duration of a single classroom interaction, the time difference from the start to the end of the interaction; the method for collecting the number of students involved in the interaction is to record the number of student accounts participating in the classroom interaction, such as the number of students who connect to the microphone, text chat, raise their hands, etc.; the method for collecting the total number of homeworks is to count the total number of homeworks completed by students within the statistical period, such as a week or a month, through the homework submission record; the method for collecting the total number of tests is to count the total number of students taking tests within the statistical period, based on the test participation record; the total number of classroom interactions The collection method is the number of classroom interactions within the statistical period, and each interaction, such as a live broadcast or an effective chat, is counted as one time; the collection method of the online learning time is to track the cumulative time students log in to the online platform, the sum of the time from login to log out, and supports statistics by day, week, and month; the collection method of the login frequency is the number of times students log in to the platform within the statistical period, such as the number of logins per day, the number of login days per week, etc.; the collection method of the resource access depth is to track the students' browsing progress on the resources, such as the percentage of video viewing, the number of document pages turned, the time spent on the courseware, etc.; the collection method of the resource relevance is determined based on the matching degree between the resource label and the course knowledge point, and is manually collected by the teacher when uploading the resource Labeling; the resource release time is collected by the timestamp of the resource being uploaded to the platform, which is automatically recorded by the system; the number of teaching resources visited is collected by counting the total number of different resources visited by students within the statistical period, with multiple visits to the same resource counted as one; the quality rating of the teaching resources is collected by sources including student ratings, such as star ratings of resources after class, and the average of ratings from multiple teachers; the update time interval of the teaching resources is collected by the system recording the time difference between two adjacent updates of the same resource; the proportion of students applicable to the teaching resources is collected by calculating based on the distribution of student grades using the resource, such as the proportion of students with passing grades to the total number of students using the resource;The difficulty coefficient of the teaching resources is collected by inferring the correct answer rate of students after using the resources. The lower the correct answer rate, the higher the difficulty coefficient. It can also be manually set by the teacher based on teaching experience. The total number of teaching resources is collected by the total number of all teaching resources stored on the platform. The total number of effective teaching resources is collected by screening out the number of resources that are actually visited by students and generate effective learning behavior, such as the number of resources with a visit depth of more than 50% and a stay time of more than 10 minutes.

[0025] S2, the teaching data preprocessing terminal processes the original teaching data set to obtain a teaching cleansing and optimization data set; First, missing values ​​in key indicators such as homework scores and test grades are filled using mean filling. Records with a missing ratio of more than 50% are directly eliminated to ensure data integrity. Then, using the Z-score and the 3σ principle, abnormal data is identified and processed, such as abnormally short homework completion times and abnormally high login frequencies. Outliers caused by recording errors are corrected to reasonable values, and real abnormal behavior data is retained and marked. Then, based on unique identifiers such as student ID, resource ID, and timestamp, duplicate records are filtered and deleted. At the same time, the formats of data from different sources are unified, such as unifying timestamps to the UTC time zone and normalizing resource difficulty coefficients to the range of [0, 1]. Finally, noise data such as invalid speeches in classroom interactions and abnormally short browsing times in resource access, such as video views of less than 10 seconds, are filtered to eliminate invalid interaction information.

[0026] S3. The AI ​​quality analysis terminal analyzes the teaching cleaning optimization data set to obtain a teaching effect impact coefficient, a student participation coefficient, and a teaching resource adaptation coefficient, and inputs the teaching effect impact coefficient, the student participation coefficient, and the teaching resource adaptation coefficient into a preset teaching quality comprehensive evaluation model to output a teaching quality comprehensive evaluation index; Furthermore, in the above technical solution, the teaching effect influence coefficient is specifically: , Among them, S i represents the score of the i-th homework, reflecting the quality of the student's completion of the single homework and is a direct reflection of the degree of knowledge mastery; W i is the weight of the i-th assignment, distinguishing the contribution of different assignments to the overall teaching effect and highlighting the evaluation value of core knowledge points; i is the completion time of the i-th operation, reflecting the impact of completion time on the operation effect. A reasonable time means a more reliable effect. max The longest job completion time in the statistical period; C j is the score of the jth test, which evaluates the comprehensive ability level of students after stage learning; T jis the length of the jth test, which measures the test complexity and student involvement. The longer the test, the wider the knowledge coverage and the higher the credibility of the results. j C is the difference between the jth test score and the previous one, reflecting the trend of score changes and the progress or regression of learning; avg is the average value of the change in grades during the statistical period; I k The effectiveness of the k-th classroom interaction is used to evaluate the promotion effect of the interactive content on the teaching objectives and filter out invalid speeches; L k is the duration of the k-th classroom interaction, reflecting the degree of students’ involvement in the discussion and constituting a complete evaluation together with the quality of interaction; N k is the number of students involved in the kth interaction, reflecting the coverage of the interaction. The more people there are, the wider the impact. n is the total number of homework, m is the total number of tests, and p is the total number of classroom interactions. It is the average of the assignment scores, a benchmark value for standardizing assignment scores and eliminating the differences in raw scores between different classes or courses; It is the average of the test scores and the benchmark value of the standardized test scores, which is used for horizontal comparison; It is the average value of classroom interaction effectiveness and the benchmark value of standardized interaction quality to avoid deviations caused by differences in the number of interactions.

[0027] Furthermore, in the above technical solution, the student participation coefficient is specifically: , in, , , , , where S1 is the online learning behavior item, S2 is the resource access quality item, D1 is the scale correction item for learning time, and D2 is the fluctuation correction item for resource depth. l F is the online learning time on the first day, which measures the time cost of students’ daily learning; l The frequency of logging in on the first day reflects the continuity and initiative of students’ learning; A r The rth resource access depth is used to evaluate students’ resource utilization, such as video viewing progress and document browsing depth; D r is the rth resource relevance, reflecting the matching degree between the resource and the current teaching content; d r The release time of the rth resource is used to measure the timeliness of the resource. Newer resources are usually more suitable for current teaching. max The longest release time of a resource within the statistical period is used to calculate the exponential decay of resource timeliness. The update frequency is set according to the resource type, such as weekly for videos and biweekly for documents. q is the number of statistical days, which determines the time range for participation evaluation, such as one week or one month. s is the number of teaching resources visited, which measures the breadth of student resource utilization. The average daily online learning time is used to standardize daily learning input for horizontal comparison; To average the depth of access to teaching resources, standardize the degree of resource utilization, and eliminate the impact of differences in resource types; A is the standard deviation of the depth of access to teaching resources, reflecting the degree of dispersion of the depth of resource access. A high degree of dispersion indicates a large difference in resource utilization. e is the Euler number, and ln is the logarithm with base e.

[0028] Furthermore, in the above technical solution, the teaching resource adaptation coefficient is specifically: , Among them, Q t Score the quality of the t-th teaching resource to evaluate the resource content quality and design level; K t is the update interval of the t-th teaching resource, which measures the timeliness of resources and avoids the use of outdated resources; H v is the proportion of students applicable to the vth teaching resource, reflecting the universality of the resource to students of different levels, such as the proportion of students who pass the exam after using the resource; M v is the difficulty coefficient of the vth teaching resource, which quantifies the difficulty of the resource to ensure that it matches the student's ability. The lower the correct answer rate, the higher the difficulty; u is the total number of teaching resources, and w is the total number of effective teaching resources; The average value of the teaching resource quality score is used to standardize the resource quality for horizontal comparison; It is the average of the proportion of applicable students, standardizing resource applicability and reflecting the overall matching degree.

[0029] Furthermore, in the above technical solution, the comprehensive teaching quality evaluation model is specifically: , Among them, E is the teaching effect influence coefficient, P is the student participation coefficient, R is the teaching resource adaptation coefficient, δ is the bias term, α, β and γ are weight coefficients, α+β+γ=1 and α, β and γ∈[0,1].

[0030] It should be noted that α, β, γ, and δ are obtained through training with historical teaching data. For the teaching data, teaching quality evaluation results, etc. within the time period, through machine learning training on these data, the optimal model parameters are determined, enabling the model to accurately output the comprehensive teaching quality index, with a value between 0 and 1. The larger the value, the higher the teaching quality. The process of machine learning training is as follows: First, collect historical teaching data including the teaching effect influence coefficient E, student participation coefficient P, teaching resource adaptation coefficient R, and the corresponding true teaching quality evaluation value, and standardize E, P, and R to eliminate the influence of dimensions. Then, define the cross-entropy loss function, and minimize the loss function under the constraints of α + β + γ = 1 and α, β, and γ ∈ [0, 1]. At initialization, randomly set α, β, and γ that satisfy this constraint, and initialize δ to 0. Then, use the chain rule to calculate the gradient, update the parameters through gradient descent, and after updating, force α + β + γ = 1 and ensure α, β, and γ ∈ [0, 1]. Repeat the process of calculating the value of QI, finding the loss function, calculating the gradient, and updating the parameters until the loss function converges to obtain the final values of α, β, γ, and δ.

[0031] S4. The teaching adjustment strategy generation terminal generates corresponding teaching adjustment strategies according to the comprehensive teaching quality index; Furthermore, in the above technical solution, the teaching adjustment strategy specifically includes: When the comprehensive teaching quality index QI ≥ m1, trigger the first-level strategy; When the comprehensive teaching quality index m2 ≤ QI < m1, trigger the second-level strategy; When the comprehensive teaching quality index QI < m2, trigger the third-level strategy.

[0032] It should be noted that m1 is initially set to 0.8, and m2 is initially set to 0.5.

[0033] Furthermore, in the above technical solution, the first-level strategy is to strengthen the teaching effect through experience replication and promotion, high-order ability cultivation, and dynamic monitoring threshold: improve student participation through incentive mechanism upgrade and precise resource push; adapt teaching resources through benchmark resource precipitation and forward-looking update; The second-level strategy is to compensate for weaknesses in the teaching effect through homework optimization, test adjustment, and enhanced interaction; improve student participation through time management intervention and resource stickiness optimization; calibrate the adaptation of teaching resources through quality and timeliness optimization and difficulty and universality adjustment; The third-level strategy is to reconstruct the teaching effect through the transformation of the teaching mode; activate student participation through personalized incentive programs and socialized learning intervention; reconstruct the adaptation of teaching resources through comprehensive resource evaluation and replacement and intelligent resource push.

[0034] It should be noted that the first-level strategy is specifically: to enhance the teaching effect, the specific measures are to extract the homework design, test mode and classroom interaction strategy of the class with high teaching effect influence coefficient value, and form a standardized template for other teachers to refer to; for △C j Approaching C avg Provide students with extended homework, advanced tests and in-depth interaction; set a fluctuation warning line for E, and if it falls below the preset threshold, the teacher will receive a message prompt; for the optimization of student participation, the specific measures are to l and F l Students with a value that is consistently higher than the average value will have privileged resources opened; based on the student's access depth A r and resource relevance D r , push personalized advanced resources; for the adaptation of teaching resources, the specific measures are to t > And H v > The resources are included in the resource library and marked with the difficulty coefficient M v and the applicable student ratio H v , for cross-class reuse; set the update time interval K according to the resource type t ; It should be noted that the preset threshold is 1.2× ; The secondary strategy is specifically: to improve the teaching effect, the specific measures are to i < Add knowledge point analysis videos for low-scoring homework; adjust homework weights i , increase the homework weight of knowledge points corresponding to frequently wrong questions, such as increasing the homework weight of weak chapters from 5% to 10%; j |>2σ C For students with j The value of σ C is the standard deviation of the test scores; k < The classroom adopts the "random roll call + group quick answer" mode, which is mandatory to cover more than 80% of the students, including is the average number of students involved in the interaction; for I k <0.5× In this case, teachers should design interactive outlines in advance, such as publishing discussion topics before class; To improve student participation, specific measures are to l < For students, push smart reminders, such as a learning pop-up window at 19:00 every day, set a mechanism to unlock privileged resources after logging in for 7 consecutive days; reduce Dr The value of A is high and r The difficulty of teaching resources with low values ​​of r >D max resources, automatically attach version update prompts to guide students to use the latest resources first; for the adaptation and calibration of teaching resources, the specific measures are to t < Resources, triggering a manual review process, are redesigned or replaced based on student feedback; t >2× The low-frequency update resources are included in the quarterly mandatory inspection list; v Less than 60% and M v High-quality resources are provided, with two versions, simple version and challenge version, allowing students to choose independently; v More than 80% and M v Low resources, marked as basic consolidation class; The three-level strategy is specifically: for the reconstruction of teaching effect, the specific measures are to reconstruct the scores of two consecutive tests C j For students with scores below 60, a special breakthrough plan will be formulated; before class, the difficulty coefficient of the teaching resources will be reduced. v 20% of the original time is used to complete the knowledge input, and the class time is used for highly interactive exercises; k <0.3× In the classroom, the mode of main teacher and AI-assisted teacher is introduced to collect students' speeches in real time, automatically filter out invalid information and guide the topic back to the teaching objectives; for the activation of student participation, the specific measures are to l Lower than And F l For students who visit less than 3 times a week, we will assign a dedicated learning consultant and formulate a tiered reward system; r and interactive participation N k , conduct mixed grouping, set group tasks, and encourage collaboration through team points; for the reconstruction of teaching resources, the specific steps are to restructure all Q t < -σ Q The process of removing from shelves, remaking and piloting will be initiated, with priority given to replacing with high-quality open source resources in the industry. Q is the standard deviation of the quality score of teaching resources; establish a dynamic feedback mechanism for resource adaptation and update H in real time v and M v For students with an accuracy rate below 50%, priority will be given to M v Low and H v High-level entry resources; push M for advanced students v High and cute t High expansion resources.

[0035] S5. The feedback execution terminal feeds back the generated teaching adjustment strategy to the relevant teaching subjects, and regularly re-evaluates the teaching quality and dynamically adjusts the strategy according to the evaluation results.

[0036] Furthermore, in the above technical solution, the dynamic adjustment process is as follows: the feedback execution terminal starts the teaching quality review mechanism according to the preset cycle, re-collects the teaching data, and obtains the new teaching quality comprehensive evaluation index QI through steps S2 and S3. new , if QI new and the comprehensive evaluation index of history teaching quality QI old If the absolute value of the difference exceeds the preset threshold, the policy adjustment logic is triggered.

[0037] Furthermore, in the above technical solution, the strategy adjustment logic includes dynamic calibration of the threshold, specifically: , , Among them, QI′ is the average value of the comprehensive evaluation index of history teaching quality, σ QI is the standard deviation of the comprehensive evaluation index of history teaching quality.

[0038] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A teaching quality control and adjustment method based on AI, characterized in that: It includes a teaching data collection terminal, a teaching data preprocessing terminal, an AI quality analysis terminal, a teaching adjustment strategy generation terminal, and a feedback execution terminal, and specifically includes the following steps: S1. The teaching data collection terminal collects data from each interaction node of the online teaching platform to obtain a teaching original data set, and the teaching original data set includes a student data set and a teaching resource data set; S2. The teaching data preprocessing terminal processes the teaching original data set to obtain a teaching cleaning and optimization data set; S3. The AI quality analysis terminal analyzes the teaching cleaning and optimization data set to obtain a teaching effect influence coefficient, a student participation coefficient, and a teaching resource adaptation coefficient, and inputs the teaching effect influence coefficient, the student participation coefficient, and the teaching resource adaptation coefficient into a preset comprehensive teaching quality evaluation model to output a comprehensive teaching quality evaluation index; S4. The teaching adjustment strategy generation terminal generates corresponding teaching adjustment strategies according to the comprehensive teaching quality index; S5. The feedback execution terminal feeds back the generated teaching adjustment strategies to relevant teaching entities, and re-evaluates the teaching quality regularly, and dynamically adjusts the strategies according to the evaluation results.

2. The AI-based teaching quality control and adjustment method according to claim 1 is characterized in that: The student data set includes the scores of homework, the time to complete homework, the test scores, the test duration, the effectiveness of classroom interaction, the duration of classroom interaction, the number of students involved in the interaction, the total number of homework, the total number of tests, the total number of classroom interactions, the online learning duration, and the login frequency; the teaching resource data set includes the depth of resource access, the resource relevance, the resource release time, the number of teaching resources accessed, the quality score of teaching resources, the update time interval of teaching resources, the applicable student ratio of teaching resources, the difficulty coefficient of teaching resources, the total number of teaching resources, and the total number of effective teaching resources.

3. The AI-based teaching quality control and adjustment method according to claim 1 is characterized in that: The teaching effect influence coefficient is specifically: ， Among them, S i To represent the score of the i-th assignment, W i is the weight of the i-th job, t i is the completion time of the i-th job, T max is the longest job completion time in the statistical period, C j is the result of the jth test, T j is the length of time for the jth test, △C j is the difference between the jth test score and the previous one, C avg is the average value of the change in grades during the statistical period, I k is the effectiveness of the k-th classroom interaction, L k is the duration of the k-th classroom interaction, N k is the number of students involved in the kth interaction, n is the total number of homework, m is the total number of tests, p is the total number of classroom interactions, is the average score of the assignments, is the average of the test scores, is the average value of classroom interaction effectiveness.

4. The AI-based teaching quality control and adjustment method according to claim 1 is characterized in that: The student participation coefficient is specifically: ， in, , , , , where S1 is the online learning behavior item, S2 is the resource access quality item, D1 is the scale correction item for learning time, and D2 is the fluctuation correction item for resource depth. l is the online learning time on day 1, F l is the login frequency on the first day, A r is the rth resource access depth, D r is the relevance of the rth resource, d r is the release time of the rth resource, D max is the longest release time of resources in the statistical period, q is the number of statistical days, s is the number of teaching resources visited, is the average daily online learning time, is the average teaching resource access depth, σ A is the standard deviation of the access depth of teaching resources, e is the Euler number, and ln is the logarithm with base e.

5. The AI-based teaching quality control and adjustment method according to claim 1 is characterized in that: The teaching resource adaptation coefficient is specifically: ， Among them, Q t is the quality score of the t-th teaching resource, K t is the update time interval of the t-th teaching resource, H v is the proportion of students applicable to the vth teaching resource, M v is the difficulty coefficient of the vth teaching resource, u is the total number of teaching resources, w is the total number of effective teaching resources, is the average value of the teaching resource quality rating, is the average of the applicable student proportions.

6. The AI-based teaching quality control and adjustment method according to claim 1 is characterized in that: The comprehensive teaching quality evaluation model is specifically: ， Among them, E is the teaching effect influence coefficient, P is the student participation coefficient, R is the teaching resource adaptation coefficient, δ is the bias term, α, β, and γ are weight coefficients, α + β + γ = 1 and α, β, and γ ∈ [0, 1].

7. The AI-based teaching quality control and adjustment method according to claim 1 is characterized in that: The teaching adjustment strategy specifically includes: When the comprehensive teaching quality index QI ≥ m1, trigger the first-level strategy; When the comprehensive teaching quality index m2 ≤ QI < m1, trigger the second-level strategy; When the comprehensive teaching quality index QI < m2, trigger the third-level strategy.

8. The AI-based teaching quality control and adjustment method according to claim 7 is characterized in that: The first-level strategy is to strengthen the teaching effect through experience replication and promotion, high-order ability cultivation, and dynamic monitoring threshold; improve student participation through incentive mechanism upgrade and resource precise push; adapt teaching resources through benchmark resource precipitation and forward-looking update; The second-level strategy is to compensate for the weakness of the teaching effect through homework optimization, test adjustment, and interaction enhancement; Improve student participation through time management intervention and resource stickiness optimization; calibrate the adaptation of teaching resources through quality and timeliness optimization and difficulty and universality adjustment; The three-level strategy is to reconstruct the teaching effect through the transformation of the teaching model; activate students' participation through personalized incentive programs and social learning interventions; and reconstruct the adaptation of teaching resources through comprehensive resource evaluation and replacement and intelligent resource push.

9. The AI-based teaching quality control and adjustment method according to claim 1 is characterized in that: The dynamic adjustment process is as follows: the feedback execution terminal starts the teaching quality review mechanism according to the preset cycle, re-collects teaching data, and obtains a new teaching quality comprehensive evaluation index QI through steps S2 and S3. new , if QI new and the comprehensive evaluation index of history teaching quality QI old If the absolute value of the difference exceeds the preset threshold, the policy adjustment logic is triggered.

10. The AI-based teaching quality control and adjustment method according to claim 9, characterized in that: The policy adjustment logic includes dynamic threshold calibration, specifically: ， Wherein, m1 is m1 as described in claim 7, m2 is m2 as described in claim 7, QI′ is the average value of the comprehensive evaluation index of history teaching quality, σ QI is the standard deviation of the comprehensive evaluation index of history teaching quality.

Images