Teaching quality dynamic evaluation system based on big data

Abstract

The invention relates to a teaching quality dynamic evaluation system based on big data, and the system comprises a data collection module which is used for obtaining an original teaching time series data flow containing student scores, classroom performance and homework quality in real time; the time sequence processing module is used for adding a time stamp to the original teaching time sequence data stream and verifying the time stamp to generate a teaching time mark data set, performing time sequence analysis on the data set to obtain a teaching fluctuation feature set, and performing multi-dimensional decomposition on the feature set to extract a teaching quality feature set; and the real-time monitoring module is used for constructing a teaching monitoring model based on the teaching target parameters, calculating a multi-dimensional deviation between the current teaching quality feature set and the historical teaching quality feature set by using the model, and generating a multi-dimensional teaching early warning signal when the deviation exceeds a preset multi-dimensional joint early warning threshold. According to the invention, real-time dynamic monitoring and risk early warning of teaching quality are realized, and evaluation efficiency and accuracy are improved.

Description

Technical Field

[0001] This invention belongs to the field of educational informatization, and in particular relates to a dynamic evaluation system for teaching quality based on big data. Background Technology

[0002] With the development of educational information technology, data-based teaching quality assessment systems have emerged. These technologies integrate student academic data to quantitatively analyze teaching effectiveness, and their key feature is the use of static indicators such as historical grades and homework completion rates to construct periodic assessment reports. Traditional technologies typically employ the following methods for teaching quality assessment: manual sampling analysis, where teachers periodically collect exam scores and homework samples, and manually calculate basic indicators such as class average and pass rate; periodic report generation, where data is summarized at the end of each semester, and static assessment reports are generated using a fixed-weight formula (e.g., 70% for grades + 30% for classroom performance); and single-dimensional early warning, where a one-way alarm is triggered only when a specific indicator (e.g., class average) falls below a preset threshold.

[0003] However, current traditional assessment methods have significant drawbacks: they are time-delayed, relying on manual sampling and end-of-semester summaries, which fail to capture real-time fluctuations in the teaching process (such as a sharp drop in classroom interaction or a continuous decline in homework quality); they are dimensionally fragmented, analyzing data such as grades, performance, and homework in isolation, ignoring the dynamic coupling effect between multi-dimensional features (such as the impact of a decline in cognitive level on behavioral dimensions); and they have a high risk of misjudgment, as fixed threshold warning mechanisms are difficult to adapt to dynamic teaching scenarios (such as the difference in indicator weights between experimental and theoretical classes), leading to missed reports or false alarms. Summary of the Invention

[0004] Therefore, it is necessary to provide a dynamic evaluation system for teaching quality based on big data that can solve the above problems.

[0005] Firstly, this application provides a dynamic evaluation system for teaching quality based on big data, including:

[0006] The data acquisition module is used to acquire raw teaching time-series data streams containing student grades, classroom performance, and homework quality in real time.

[0007] The time series processing module is used to add time stamps to the original teaching time series data stream and verify the timestamps to generate a teaching time stamp dataset; perform time series analysis on the teaching time stamp dataset to obtain a teaching fluctuation feature set; and perform multi-dimensional decomposition on the teaching fluctuation feature set to extract a teaching quality feature set.

[0008] The real-time monitoring module is used to build a teaching monitoring model based on teaching objective parameters;

[0009] Using a teaching monitoring model, the multi-dimensional deviation between the teaching quality feature set and the historical teaching quality feature set is calculated; when the multi-dimensional deviation exceeds the preset multi-dimensional joint early warning threshold, a multi-dimensional teaching early warning signal is generated.

[0010] In one embodiment, the timing processing module is further configured to:

[0011] Add a time stamp to each data unit in the original teaching time-series data stream to generate a primary time-stamped data stream;

[0012] Taking the primary time-stamped data stream as input, the system uses a dual-channel verification mechanism to check the time stamp against the physical timestamp at the data acquisition end, and outputs a time-stamped dataset that has passed the verification.

[0013] Based on a preset teaching cycle template, the corresponding time-stamped datasets are categorized into teaching stages according to the time stamps of the validated time-stamped datasets, generating teaching time-stamped datasets containing teaching stage labels.

[0014] In one embodiment, the timing processing module is further configured to:

[0015] Based on the teaching stage labels in the teaching time-stamped dataset, the teaching time-stamped dataset is grouped, and teaching time-stamped data is extracted from different groups using preset grouping sampling rules to form time-stamped subsets;

[0016] The fluctuation feature vector of teaching fluctuation is obtained by performing fluctuation feature extraction on each time scale subset using the change point detection algorithm;

[0017] A sensitivity analysis model is used to calculate the correlation of teaching fluctuation feature vectors between different teaching stages, and a cross-stage correlation feature matrix is ​​generated.

[0018] By integrating the teaching fluctuation feature vector with the cross-stage correlation feature matrix, a teaching fluctuation feature set is generated.

[0019] In one embodiment, the timing processing module is further configured to:

[0020] The teaching fluctuation feature set is decomposed in parallel according to the knowledge dimension, behavior dimension and cognition dimension to generate corresponding dimensional feature subsets;

[0021] Calculate the statistical correlation measure between each pair of the knowledge dimension feature subset, the behavior dimension feature subset, and the cognitive dimension feature subset, and generate the inter-dimensional correlation matrix;

[0022] By generating a correlation matrix between dimensions, a model of dimensional coupling strength coefficients is constructed through a pre-defined weighted fusion algorithm.

[0023] By integrating the inter-dimensional correlation matrix and the dimensional coupling strength coefficient model, a set of teaching quality features is generated.

[0024] In one embodiment, the real-time monitoring module is further configured to:

[0025] Based on a pre-set library of teaching scenario configuration parameters, in response to multi-dimensional teaching warning signals or preset teaching stage conversion instructions, the configuration parameter set corresponding to the current teaching scenario identifier is matched.

[0026] Adjust the weight configuration of each dimension in the teaching monitoring model according to the configuration parameter set;

[0027] Using the adjusted weights for each dimension, the scenario-adaptive multidimensional deviation values ​​between the teaching quality feature set and the historical teaching quality feature set are calculated.

[0028] When the scenario-adaptive multi-dimensional deviation value is detected to converge to the preset steady-state range or when a teaching stage transition is triggered, a teaching status assessment report is generated by calling the report template based on the scenario-adaptive multi-dimensional deviation value and the current weight configuration of each dimension.

[0029] In one embodiment, after the real-time monitoring module generates a teaching quality feature set, it calculates a multi-dimensional score for the teaching quality feature set using the following formula:

[0030]

[0031] Where d is the dimension classification label, k is the knowledge dimension, b is the behavior dimension, c is the cognitive dimension, i and j are the interaction identifiers between different dimensions, and f is the interaction identifier between different dimensions. d For a subset of dimensional features, φ represents the elements of the inter-dimensional correlation matrix. ij ρ is the dimensional coupling strength coefficient, α is the preset coupling effect weight factor, and ρ is the multi-dimensional score of the teaching quality feature set.

[0032] In one embodiment, the real-time monitoring module is further configured to:

[0033] In response to the received request for obtaining the teaching status assessment report, a teaching quality feature set update instruction is sent to the time-series processing module;

[0034] Based on the updated teaching quality feature set returned by the time series processing module, a new teaching status assessment report is generated.

[0035] Secondly, this application also provides a method for dynamic evaluation of teaching quality based on big data, including:

[0036] Real-time acquisition of raw teaching time-series data streams including student grades, classroom performance, and homework quality;

[0037] Add time stamps to the original teaching time-series data stream and verify the timestamps to generate a teaching time-stamped dataset;

[0038] Time series analysis was performed on the teaching time-scale dataset to obtain a set of teaching fluctuation characteristics;

[0039] The teaching fluctuation feature set is decomposed in multiple dimensions to extract the teaching quality feature set;

[0040] Based on teaching objective parameters, a teaching monitoring model is constructed;

[0041] Using a teaching monitoring model, we calculate the multi-dimensional deviation between the teaching quality feature set and the historical teaching quality feature set.

[0042] When the multi-dimensional deviation exceeds the preset multi-dimensional joint early warning threshold, a multi-dimensional teaching early warning signal is generated.

[0043] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to realize the functions of the above-mentioned dynamic evaluation system for teaching quality based on big data.

[0044] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the functions of the aforementioned big data-based dynamic evaluation system for teaching quality.

[0045] The aforementioned big data-based dynamic evaluation system for teaching quality acquires raw, time-series teaching data streams containing student grades, classroom performance, and homework quality in real time through a data acquisition module. This ensures the immediacy and continuity of teaching data, avoiding delays caused by manual sampling and end-of-semester summarization. It can capture real-time fluctuations in the teaching process (such as a sudden drop in classroom interaction or a continuous decline in homework quality). The time-series processing module adds time stamps to the raw data stream and verifies the timestamps to generate a teaching time-stamped dataset. Through time-series analysis, it extracts a set of teaching fluctuation features and performs multi-dimensional decomposition to generate a set of teaching quality features, addressing the shortcomings of fragmented dimensions and breaking away from traditional isolated approaches. By analyzing the limitations of grades, performance, and assignment data, this approach dynamically captures the coupling effects between knowledge, behavior, and cognitive dimensions (such as the impact of declining cognitive levels on behavior) through multi-dimensional feature fusion, avoiding misjudgments caused by ignoring multi-dimensional interactions. The real-time monitoring module constructs a teaching monitoring model based on teaching objective parameters, calculates the multi-dimensional deviation between the current teaching quality feature set and the historical dataset, and generates an early warning signal when the deviation exceeds a preset multi-dimensional joint early warning threshold. This eliminates the high risk of misjudgment (such as missed reports or false alarms) caused by fixed threshold mechanisms, achieving real-time dynamic monitoring and risk warning of teaching quality, and significantly improving evaluation efficiency and accuracy. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a structural diagram of a dynamic evaluation system for teaching quality based on big data, according to the present invention.

[0048] Figure 2 This is a flowchart of a dynamic evaluation method for teaching quality based on big data, according to the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] In one embodiment, such as Figure 1As shown, a dynamic evaluation system for teaching quality based on big data is provided. This embodiment illustrates the system deployed on a terminal, but it is understood that the system can also be deployed on a server and can be applied to an architecture that includes both terminals and servers, and is implemented through the interaction between the terminal and the server. In the terminal deployment mode, the teacher's terminal device (such as a smart teaching tablet) can capture classroom performance data streams in real time through the data acquisition module, and simultaneously establish a secure connection with the campus database to obtain time-series data on student grades and homework quality. When a server architecture is adopted, the cloud computing node performs time-series analysis and multi-dimensional decomposition operations in the time-series processing module, and handles concurrent data streams from multiple classrooms through a load balancing strategy. In the terminal-server hybrid architecture, the edge computing terminal performs primary time-stamped data stream generation and timestamp verification, while the cloud server cluster undertakes the tasks of extracting teaching fluctuation feature sets and constructing dimensional coupling strength coefficient models. The two sides achieve data synchronization and command transmission through encrypted API interfaces. Its typical application scenario is as follows: When a coupled fluctuation occurs between student cognitive dimension indicators (such as delayed response to classroom questions) and behavioral dimensions (such as frequency of use of interactive devices) during theoretical teaching, the terminal sensor collects abnormal data streams in real time. The server-side time-series processing module captures cross-dimensional correlation features through a change-point detection algorithm, and the real-time monitoring module then calculates the scenario-adaptive multi-dimensional deviation value, triggering a teaching warning signal to the teacher's terminal device. During experimental class scenario switching, the system automatically calls the teaching scenario configuration parameter library to dynamically adjust the monitoring model weights, increasing the monitoring sensitivity of knowledge dimensions (such as operational standardization) by 40%. Simultaneously, a visual teaching status evaluation report can be generated through the terminal device. In this embodiment, the system includes:

[0051] The data acquisition module 101 is used to acquire raw teaching time-series data streams containing student grades, classroom performance, and homework quality in real time.

[0052] The data acquisition module 101 is used to acquire raw teaching time-series data streams containing student grades, classroom performance, and homework quality in real time. It can achieve full-cycle data capture through a distributed terminal architecture. This includes student grade data (a numerical sequence composed of periodic assessment results and in-class quiz scores); classroom performance data (covering unstructured behavioral indicators such as teacher-student interaction frequency, attention concentration index, and device operation logs); and homework quality data (represented as a multi-dimensional set of values ​​including completion rate, accuracy rate, and innovation score). The module's implementation involves three technical layers: a bottom-layer data perception layer deployed on teaching terminals periodically captures classroom performance signals; a middle-layer data integration layer uses edge computing nodes to perform time alignment operations on heterogeneous data, converting discrete homework submission events into a continuous stream of quality indicators through a sliding window mechanism; and an upper-layer data output layer built on a streaming processing engine to generate structured triples (timestamp, data type, data payload) conforming to the ISO / IEC 15444 standard.

[0053] The time series processing module 102 is used to add time stamps to the original teaching time series data stream and verify the timestamps to generate a teaching time stamp dataset; perform time series analysis on the teaching time stamp dataset to obtain a teaching fluctuation feature set; and perform multi-dimensional decomposition on the teaching fluctuation feature set to extract a teaching quality feature set.

[0054] The dataset includes: a teaching time-stamped dataset (a structured dataset whose timestamp validity is verified through a dual-channel validation mechanism, where each data unit is appended with a time stamp and categorized into teaching stage labels based on a preset teaching cycle template); a teaching fluctuation feature set (a collection of dynamically changing fluctuation patterns during the teaching process, generated by dividing the teaching time-stamped dataset into time-stamped subsets using group sampling rules, extracting feature vectors using a change point detection algorithm, and then combining them with a sensitivity analysis model to calculate a cross-stage correlation feature matrix); and a teaching quality feature set (a coupled feature model of knowledge dimension (covering knowledge point mastery), behavioral dimension (such as interaction frequency), and cognitive dimension (such as thinking response delay), achieved through parallel decomposition of dimensional feature subsets, calculation of inter-dimensional correlation matrices, construction of dimensional coupling strength coefficient models, and weighted fusion); and data input. The first layer receives the raw teaching time-series data stream, performs time stamp appending operations through edge computing nodes to generate a primary time-stamped data stream, compares the physical timestamps of the data acquisition end with a dual-channel verification mechanism to eliminate clock drift, and outputs the verified time-stamped dataset. Then, it automatically adds teaching stage labels according to the teaching cycle template to form the final teaching time-stamped dataset. In the analysis layer, a windowed grouping strategy is used to extract a subset of time-stamped data. A change point detection algorithm based on Bayesian inference is applied to identify abnormal fluctuation points and generate teaching fluctuation feature vectors. Sensitivity analysis is used to quantify the correlation of features in different teaching stages to generate a cross-stage correlation feature matrix. In the feature extraction layer, the fluctuation feature set is decomposed into three dimensions and the covariance matrix is ​​calculated to generate an inter-dimensional correlation matrix. A dimensional coupling strength coefficient model is constructed using a preset weighting algorithm and fused to output the teaching quality feature set.

[0055] The real-time monitoring module 103 is used to construct a teaching monitoring model based on teaching objective parameters; use the teaching monitoring model to calculate the multi-dimensional deviation between the teaching quality feature set and the historical teaching quality feature set; and generate a multi-dimensional teaching early warning signal when the multi-dimensional deviation exceeds the preset multi-dimensional joint early warning threshold.

[0056] Among them, the teaching objective parameters (a dynamic weight configuration set pre-installed in the teaching scenario configuration parameter library (such as the difference in knowledge dimension weights between theoretical and experimental courses), used to adapt to different teaching scenarios); the teaching monitoring model (represented as a computational framework built on a multi-dimensional coupling strength coefficient model, dynamically adjusting the deviation calculation logic through dimension weight configuration); multi-dimensional deviation (analyzed as the comprehensive offset value of the current teaching quality feature set (including subsets of knowledge dimension features, subsets of behavioral dimension features, and subsets of cognitive dimension features) and the historical dataset in terms of statistical correlation measurement); multi-dimensional teaching early warning signals (cross-dimensional alarm instructions triggered when the offset value exceeds the joint threshold, which can directly drive the teacher's terminal device to intervene); and the input layer. The system receives teaching objective parameters and teaching quality feature sets. During the model building phase, it calls a pre-configured parameter library to match the current teaching scenario identifier (such as responding to early warning signals or stage transition instructions) and adjusts the weight configuration of each dimension in the teaching monitoring model to eliminate the risk of misjudgment caused by fixed thresholds. In the deviation calculation layer, a scenario-adaptive algorithm is applied to calculate the deviation value between the teaching quality feature set and the historical dataset under the multi-dimensional coupling effect. In the early warning generation layer, the deviation value is monitored in real time. When it exceeds the preset multi-dimensional joint early warning threshold (such as the coupling offset between the cognitive dimension and the behavioral dimension exceeds the standard), a multi-dimensional teaching early warning signal is generated and output to the monitoring terminal. At the same time, when the deviation value converges to the preset steady-state range, the teaching status evaluation report generation mechanism is triggered.

[0057] The aforementioned big data-based dynamic evaluation system for teaching quality acquires raw, time-series teaching data streams containing student grades, classroom performance, and homework quality in real time through a data acquisition module. This directly overcomes the time lag inherent in previous technologies. Employing a distributed terminal architecture, it continuously captures data streams (such as events like sudden drops in classroom attention index), avoiding minute-level delays associated with traditional manual sampling and end-of-semester aggregation. A time-series processing module adds time stamps to the raw data streams and verifies the timestamps using a dual-channel verification mechanism, generating a teaching time-stamped dataset labeled with teaching stages. Time-series analysis (such as change point detection algorithms) extracts teaching fluctuation feature sets and performs multi-dimensional decomposition (parallel processing according to knowledge, behavior, and cognition dimensions), resolving the problem of dimensional fragmentation. Breaking away from the limitations of isolated analysis of grades, performance, and assignment data, this approach calculates statistical correlations between dimensions and constructs a dimensional coupling strength coefficient model (such as the dynamic correlation between cognitive dimension response delay and behavioral dimension interaction frequency). This eliminates the risk of misjudgment caused by neglecting multidimensional interactions in traditional assessments. A real-time monitoring module builds a teaching monitoring model based on teaching objective parameters (such as the weight difference between experimental and theoretical classes), calculating the multidimensional deviation between the current teaching quality feature set and historical datasets (quantifying cross-dimensional coupling effects). When the deviation exceeds a preset multidimensional joint warning threshold, a warning signal is generated, addressing the high risk of misjudgment caused by fixed threshold mechanisms. For example, a precise alarm is triggered by the coupling shift between declining operational standardization in experimental classes and grade fluctuations, avoiding false alarms or missed reports.

[0058] In one embodiment, the timing processing module 102 is further configured to:

[0059] Add a time stamp to each data unit in the original teaching time-series data stream to generate a primary time-stamped data stream;

[0060] Taking the primary time-stamped data stream as input, the system uses a dual-channel verification mechanism to check the time stamp against the physical timestamp at the data acquisition end, and outputs a time-stamped dataset that has passed the verification.

[0061] Based on a preset teaching cycle template, the corresponding time-stamped datasets are categorized into teaching stages according to the time stamps of the validated time-stamped datasets, generating teaching time-stamped datasets containing teaching stage labels.

[0062] Specifically, the process involves: a primary time-stamped data stream (an initial structured dataset carrying local application-layer timestamps); inputting this primary time-stamped data stream and verifying the timestamps against the physical timestamps of the data acquisition terminal using a dual-channel verification mechanism (parallel comparison of the hardware clock channel and the application-layer clock channel), outputting a verified time-stamped dataset (a set of data that passes CRC verification); based on a preset teaching cycle template (containing standardized stage labels such as new lesson introduction, core explanation, and experimental operation), classifying the corresponding datasets into teaching stages according to the timestamps of the verified time-stamped datasets, and generating a teaching time-stamped dataset containing teaching stage labels (represented as a time-series data matrix with normalized stage identifiers). In the data labeling layer, edge computing nodes periodically add time stamps to raw data (such as a single classroom Q&A record) to form a primary time-stamped data stream. In the verification layer, a dual-channel time synchronization protocol is used. The hardware channel reads the physical clock of the sensor, and the application channel parses the system clock. Clock drift is eliminated by comparison through a sliding window, and only data units that pass the verification are output. In the classification layer, the teaching cycle template maps timestamps to teaching stages (such as 9:00-9:15 being automatically marked as the introduction of a new lesson). The mapping relationship between timestamps and stage labels can be established through hash indexes to generate a teaching time-stamped dataset with stage labels.

[0063] In one embodiment, the timing processing module 102 is further configured to:

[0064] Based on the teaching stage labels in the teaching time-stamped dataset, the teaching time-stamped dataset is grouped, and teaching time-stamped data is extracted from different groups using preset grouping sampling rules to form time-stamped subsets;

[0065] The fluctuation feature vector of teaching fluctuation is obtained by performing fluctuation feature extraction on each time scale subset using the change point detection algorithm;

[0066] A sensitivity analysis model is used to calculate the correlation of teaching fluctuation feature vectors between different teaching stages, and a cross-stage correlation feature matrix is ​​generated.

[0067] By integrating the teaching fluctuation feature vector with the cross-stage correlation feature matrix, a teaching fluctuation feature set is generated.

[0068] For example, based on the teaching stage labels (such as predefined stage identifiers like new lesson introduction, core explanation, and experimental operation) in the teaching time-stamped dataset, the dataset is grouped (dividing the dataset into independent data groups according to stage labels). Using pre-defined grouping sampling rules (such as a uniform sampling strategy to ensure consistent sampling ratios for each group), teaching time-stamped data is extracted from different groups to form time-stamped subsets (time-series data subsets representing specific teaching stages). A change point detection algorithm (such as a Bayesian inference-based algorithm model) is used to extract fluctuation features from each time-stamped subset, identifying anomalous abrupt changes in the data stream (such as a sudden drop in classroom interaction frequency), generating a teaching fluctuation feature vector (represented as a multi-dimensional feature array containing fluctuation amplitude, frequency, and duration). A sensitivity analysis model (such as the Pearson correlation coefficient calculation framework) is used to calculate the correlation between teaching fluctuation feature vectors across different teaching stages (such as the fluctuation correlation between the new lesson introduction stage and the experimental operation stage), generating a cross-stage correlation feature matrix (a mathematical matrix quantifying the dynamic dependencies between stages). Finally, the teaching fluctuation feature vectors are fused. The system generates a teaching fluctuation feature set (parsed into a feature set covering micro-fluctuation patterns and cross-stage correlations) by combining a cross-stage correlation feature matrix (integrating vectors and matrix elements through a weighted average algorithm). In implementation, the data input layer loads a teaching time-scaled dataset labeled with teaching stages. During the grouping stage, hash indexing technology is applied to divide data into groups by label (e.g., data from the core explanation stage is grouped independently). The sampling rule uses a sliding window mechanism to extract data points at fixed intervals (e.g., every 5 minutes) to ensure that the time-scaled subsets cover representative samples from each stage. In the feature extraction layer, a change point detection algorithm performs real-time scanning on each time-scaled subset (e.g., detecting abnormal declines in homework submission rates within 10 seconds), outputting a teaching fluctuation feature vector through a probability model (vector dimensions include fluctuation point location, intensity value, etc.). The sensitivity analysis model calculates the inter-stage feature correlation (e.g., analyzing the covariance of fluctuation vectors between theoretical and experimental lessons), generating a cross-stage correlation feature matrix (matrix elements store correlation coefficients). The fusion layer calls a linear weighting function (weights are assigned by preset rules) to integrate vectors and matrices, outputting the teaching fluctuation feature set.

[0069] In one embodiment, the timing processing module 102 is further configured to:

[0070] The teaching fluctuation feature set is decomposed in parallel according to the knowledge dimension, behavior dimension and cognition dimension to generate corresponding dimensional feature subsets;

[0071] Calculate the statistical correlation measure between each pair of the knowledge dimension feature subset, the behavior dimension feature subset, and the cognitive dimension feature subset, and generate the inter-dimensional correlation matrix;

[0072] By generating a correlation matrix between dimensions, a model of dimensional coupling strength coefficients is constructed through a pre-defined weighted fusion algorithm.

[0073] By integrating the inter-dimensional correlation matrix and the dimensional coupling strength coefficient model, a set of teaching quality features is generated.

[0074] Specifically, the teaching fluctuation feature set is decomposed in parallel according to the knowledge dimension (covering knowledge point mastery indicators, such as grade fluctuation data), the behavioral dimension (such as teacher-student interaction frequency, equipment operation logs), and the cognitive dimension (such as thinking response delay, question and answer response time). (The feature set is divided into independent data blocks using a distributed computing framework) to generate corresponding dimensional feature subsets (i.e., knowledge dimension feature subset, behavioral dimension feature subset, and cognitive dimension feature subset, each subset representing a feature vector of a single dimension); the statistical correlation between each pair of the knowledge dimension feature subset, behavioral dimension feature subset, and cognitive dimension feature subset is calculated. The system measures (using Pearson correlation coefficient or covariance calculation framework to analyze the interaction relationship between subsets) to generate an inter-dimensional correlation matrix (a mathematical matrix storing the correlation coefficients between knowledge-behavior, knowledge-cognition, and behavior-cognition dimensions); using this generated inter-dimensional correlation matrix, a dimensional coupling strength coefficient model is constructed through a preset weighted fusion algorithm (a linear weighted model trained on historical data); the inter-dimensional correlation matrix and the dimensional coupling strength coefficient model are fused (through the summation of the product of matrix elements and coupling coefficients) to generate a teaching quality feature set (a comprehensive index set of multi-dimensional coupling features, used as input for the real-time monitoring module). In implementation, the data input layer loads the teaching fluctuation feature set. In the decomposition layer, a parallel processing engine (such as Apache Spark) is invoked to perform feature segmentation according to predefined dimension mapping rules (such as associating knowledge dimension with academic performance data, behavioral dimension with sensor logs, and cognitive dimension with response latency), generating dimensional feature subsets. In the matrix generation layer, the statistical calculation module performs pairwise correlation analysis on the subsets (e.g., calculating the covariance between the knowledge dimension subset and the behavioral dimension subset), outputting the inter-dimensional correlation matrix. In the model building layer, a weighted fusion algorithm (with weight factors determined by the machine learning model based on historical datasets) processes the matrix elements to construct a dimensional coupling strength coefficient model. In the feature output layer, the matrix and model are integrated through an element-level fusion function to generate the teaching quality feature set.

[0075] In one embodiment, the real-time monitoring module 103 is further configured to:

[0076] Based on a pre-set library of teaching scenario configuration parameters, in response to multi-dimensional teaching warning signals or preset teaching stage conversion instructions, the configuration parameter set corresponding to the current teaching scenario identifier is matched.

[0077] Adjust the weight configuration of each dimension in the teaching monitoring model according to the configuration parameter set;

[0078] Using the adjusted weights for each dimension, the scenario-adaptive multidimensional deviation values ​​between the teaching quality feature set and the historical teaching quality feature set are calculated.

[0079] When the scenario-adaptive multi-dimensional deviation value is detected to converge to the preset steady-state range or when a teaching stage transition is triggered, a teaching status assessment report is generated by calling the report template based on the scenario-adaptive multi-dimensional deviation value and the current weight configuration of each dimension.

[0080] For example, based on a pre-set teaching scenario configuration parameter library (storing dimensional weight configuration sets for different scenarios such as theoretical classes and experimental classes), in response to multi-dimensional teaching early warning signals (such as alarms for excessive coupling deviation between cognitive and behavioral dimensions) or preset teaching stage conversion instructions (such as trigger signals for switching from theoretical classes to experimental classes), the configuration parameter set (containing a dynamic parameter matrix of knowledge dimension weights, behavioral dimension weights, and cognitive dimension weights) corresponding to the current teaching scenario identifier is matched through a scenario matching engine; the weight configuration of each dimension in the teaching monitoring model is dynamically adjusted according to this configuration parameter set (for example, increasing the knowledge dimension weight by 40% in the experimental class scenario). The behavioral dimension weight is reduced by 20%. Using the adjusted dimension weight configuration, the teaching quality feature set (including subsets of knowledge dimension features, behavioral dimension features, and cognitive dimension features) and the historical teaching quality feature set are recalculated under the multi-dimensional coupling effect to determine the scenario-adaptive multi-dimensional deviation value. When the deviation value is detected to converge to the preset steady-state range (e.g., fluctuation range < ±5% for 5 consecutive minutes) or a new teaching stage transition is triggered, a teaching status evaluation report (e.g., a visual document including a three-dimensional coupling heatmap and stage transition trend analysis) is generated based on the current deviation value and weight configuration by calling the report template (built-in HTML dynamic rendering engine). In implementation, after receiving a warning signal or stage transition instruction, the scene matching layer extracts the parameter set corresponding to the scene identifier from the configuration parameter library through hash retrieval (e.g., the experimental class scene identifier is mapped to a knowledge dimension weight of 0.6, a behavior dimension weight of 0.2, and a cognitive dimension weight of 0.2). The weight adjustment layer overwrites the weight register of the teaching monitoring model in real time, and adopts a double buffering mechanism to ensure uninterrupted weight switching. The deviation calculation layer executes the scene-adaptive deviation formula in the streaming computing engine, where the historical dataset is dynamically updated using a sliding window mechanism (the window size is determined by the duration of the teaching stage), and the covariance calculation introduces an adjusted coupling coefficient. The report generation layer monitors the rate of change of deviation values ​​through the steady-state detection module. When the convergence condition is met (e.g., the standard deviation is continuously lower than the threshold) or a stage end signal is received, the JSON template engine is called to fill in the deviation data and weight configuration, and outputs an evaluation report in PDF / HTML format.

[0081] In one embodiment, after the real-time monitoring module 103 generates a teaching quality feature set, it calculates a multi-dimensional score for the teaching quality feature set using the following formula:

[0082]

[0083] Where d is the dimension classification label, k is the knowledge dimension, b is the behavior dimension, c is the cognitive dimension, i and j are the interaction identifiers between different dimensions, and f is the interaction identifier between different dimensions. d For a subset of dimensional features, φ represents the elements of the inter-dimensional correlation matrix. ijρ is the dimensional coupling strength coefficient, α is the preset coupling effect weight factor, and ρ is the multi-dimensional score of the teaching quality feature set.

[0084] Specifically, the multi-dimensional score ρ (a scalar value quantifying the overall quality of the knowledge dimension (k), behavioral dimension (b), and cognitive dimension (c)) and the dimensional feature subset f d (The knowledge dimension feature subsets extracted through the time-series processing module, such as knowledge point mastery indicators; the behavioral dimension feature subsets, such as teacher-student interaction frequency; and the cognitive dimension feature subsets, such as thinking response delay), elements of the inter-dimensional correlation matrix. (The variance measure stored in the inter-dimensional correlation matrix quantifies the data dispersion within a dimension), dimensional coupling strength coefficient φ ij (Model coefficients constructed using a pre-defined weighted fusion algorithm represent the dynamic coupling strength between knowledge-behavior, knowledge-cognition, and behavior-cognition dimensions). A pre-defined coupling effect weight factor α (configurable parameter, default value 0.5, used to adjust the contribution weight of the coupling effect in the total score) is used. Dimension classification labels d identify dimension classifications (k for knowledge dimension, b for behavior dimension, c for cognition dimension). Interaction identifiers (i and j) are pairing indices between different dimensions (e.g., i represents the knowledge dimension, j represents the behavior dimension). Input the teaching quality feature set (containing subsets of dimensional features and the correlation matrix between dimensions), and perform parallel processing in the computation engine. The first operation is a summation. Calculate the normalized energy value of each feature subset in each dimension (using the squared L2 norm ||f) d || 2 Divide by variance (Eliminating intra-dimensional fluctuations), the second summation operation α·∑ i≠j φ ij ·cov(f i f j Introduce a weighting factor α and calculate the inter-dimensional covariance cov(f) i f j ) and coupling coefficient φ ij The sum of the products (capturing the interaction effect between the knowledge dimension and the behavior dimension, such as the dynamic correlation between declining grades and decreasing interaction frequency) is used to output the ρ value as a multi-dimensional score.

[0085] In one embodiment, the real-time monitoring module 103 is further configured to:

[0086] In response to the received request for obtaining the teaching status assessment report, a teaching quality feature set update instruction is sent to the timing processing module 102;

[0087] Based on the updated teaching quality feature set returned by the time-series processing module 102, a new teaching status assessment report is generated.

[0088] For example, in response to a received request to obtain a teaching status assessment report (either a report generation instruction actively triggered by the user through the teacher's terminal or an automatic request triggered by a system timed task), a teaching quality feature set update instruction (a set of commands containing time range identifiers and data resampling parameters) is sent to the time series processing module 102; based on the updated teaching quality feature set returned by the time series processing module 102 (characterized as multi-dimensional coupled data containing teaching fluctuation characteristics within the latest time window), the teaching status assessment report can be regenerated through the streaming rendering engine (the output is a visual document containing a three-dimensional coupled heatmap and real-time deviation trends). In implementation, when the HTTP request interface detects a report retrieval instruction (e.g., a teacher clicking the generate report button), the instruction parser extracts the request parameters (e.g., the time range of the most recent hour) and generates an update instruction with a timestamp. After receiving the instruction, the time series processing module 102 dynamically adjusts the data sampling window (e.g., extracting the most recent 60-minute teaching time-stamped dataset), re-executes the fluctuation feature extraction and multi-dimensional decomposition (the specific process is described in claims 3-4), and returns the updated feature set. The real-time monitoring module 103 calls the report template (with a built-in HTML dynamic rendering engine) and applies the dimensional coupling strength coefficient φ from the updated feature set. ij The multi-dimensional scores ρ are populated in real time into visual charts (such as a heatmap linking knowledge and behavior dimensions), and the scenario-adaptive deviation value change curve is written synchronously. Finally, a PDF / HTML format report is output.

[0089] The aforementioned big data-based dynamic evaluation system for teaching quality acquires raw, time-series teaching data streams containing student grades, classroom performance, and homework quality in real time through a data acquisition module. It employs a distributed terminal architecture to capture micro-fluctuations throughout the entire teaching cycle (such as sudden drops in classroom attention index) at millisecond-level sampling frequencies, solving the problem of time lag and overcoming the delays of traditional manual sampling and end-of-semester summarization, thus achieving low-latency anomaly detection. The time-series processing module adds time stamps to the data stream and verifies the timestamps, generating a teaching time-stamped dataset labeled with teaching stages. Then, a change point detection algorithm is used to extract a set of teaching fluctuation features, which are then decomposed in parallel according to knowledge, behavior, and cognitive dimensions. The statistical correlation between dimensions is calculated, and a dimensional coupling strength coefficient model φ is constructed. ij This approach eliminates the limitations of dimensional fragmentation, breaks through the isolation of analyzing grades, performance, and assignments, and dynamically captures the real-time impact of declining cognitive levels on behavioral dimensions (such as the coupling effect of delayed question-and-answer responses leading to decreased interaction frequency), reducing the risk of misjudgment. A teaching monitoring model is constructed based on teaching objective parameters (such as the weighting configuration of experimental classes) through a real-time monitoring module, utilizing formulas... The system calculates a multi-dimensional score ρ. When the scene adaptation deviation exceeds the joint early warning threshold, an early warning signal is generated. Combined with a dynamic feature set update mechanism, the model parameters are automatically adjusted to fundamentally address the high risk of misjudgment in background technology—eliminating false alarms caused by the difference in fixed thresholds between experimental and theoretical lessons (e.g., accurately triggering alarms when operational standardization declines in experimental lessons). This ultimately forms a closed-loop monitoring chain: low-latency data acquisition → multi-dimensional coupled feature fusion → dynamic scene adaptation early warning, resolving issues of time lag, dimensional fragmentation, and high risk of misjudgment.

[0090] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0091] Based on the same inventive concept, this application also provides a method for dynamic evaluation of teaching quality based on big data to implement the aforementioned dynamic evaluation system for teaching quality based on big data. The solution provided by this method is similar to the implementation scheme described above. Therefore, the specific limitations of one or more embodiments of the dynamic evaluation method for teaching quality based on big data provided below can be found in the limitations of the dynamic evaluation system for teaching quality based on big data described above, and will not be repeated here.

[0092] In one exemplary embodiment, such as Figure 2 As shown, a dynamic evaluation method for teaching quality based on big data is provided, including:

[0093] S01, real-time acquisition of raw teaching time-series data streams including student grades, classroom performance, and homework quality;

[0094] S02, add time stamps to the original teaching time series data stream and verify the timestamps to generate a teaching time stamp dataset;

[0095] S03, perform time series analysis on the teaching timescale dataset to obtain the teaching fluctuation feature set;

[0096] S04, decompose the teaching fluctuation feature set in multiple dimensions and extract the teaching quality feature set;

[0097] S05, Construct a teaching monitoring model based on teaching objective parameters;

[0098] S06, using a teaching monitoring model, calculate the multi-dimensional deviation between the teaching quality feature set and the historical teaching quality feature set;

[0099] S07: When the multi-dimensional deviation exceeds the preset multi-dimensional joint early warning threshold, a multi-dimensional teaching early warning signal is generated.

[0100] In one embodiment, a timestamp is appended to the original teaching time-series data stream and the timestamp is verified to generate a teaching time-stamped dataset, including:

[0101] S11 adds a time stamp to each data unit in the original teaching time-series data stream to generate a primary time-stamped data stream;

[0102] S12 takes the primary time-stamped data stream as input, verifies the time stamp against the physical timestamp of the data acquisition end through a dual-channel verification mechanism, and outputs the time-stamped dataset that has passed the verification.

[0103] S13, based on the preset teaching cycle template, classifies the corresponding time-stamped datasets into teaching stages according to the time stamps of the verified time-stamped datasets, and generates a teaching time-stamped dataset containing teaching stage labels.

[0104] In one embodiment, time series analysis is performed on the teaching time-stamped dataset to obtain a set of teaching fluctuation characteristics, including:

[0105] S21. Based on the teaching stage labels in the teaching time-stamp dataset, the teaching time-stamp dataset is grouped, and teaching time-stamp data is extracted from different groups using preset grouping sampling rules to form time-stamp subsets.

[0106] S22, the change point detection algorithm is used to extract fluctuation features for each time scale subset to obtain the teaching fluctuation feature vector;

[0107] S23, use a sensitivity analysis model to calculate the correlation of teaching fluctuation feature vectors between different teaching stages, and generate a cross-stage correlation feature matrix;

[0108] S24, integrates the teaching fluctuation feature vector with the cross-stage correlation feature matrix to generate a teaching fluctuation feature set.

[0109] In one embodiment, the teaching fluctuation feature set is decomposed into multiple dimensions to extract the teaching quality feature set, including:

[0110] S31, decompose the teaching fluctuation feature set in parallel according to the knowledge dimension, behavior dimension and cognition dimension to generate the corresponding dimensional feature subsets;

[0111] S32 calculates the statistical correlation measure between each pair of the knowledge dimension feature subset, the behavior dimension feature subset, and the cognitive dimension feature subset, and generates the inter-dimensional correlation matrix;

[0112] S33, by generating a correlation matrix between dimensions, a dimensional coupling strength coefficient model is constructed through a preset weighted fusion algorithm;

[0113] S34 integrates the inter-dimensional correlation matrix and the dimensional coupling strength coefficient model to generate a teaching quality feature set.

[0114] In one embodiment, the method further includes:

[0115] S41, based on a pre-set teaching scenario configuration parameter library, responds to multi-dimensional teaching warning signals or preset teaching stage conversion instructions, and matches the configuration parameter set corresponding to the current teaching scenario identifier;

[0116] S42, adjust the weight configuration of each dimension in the teaching monitoring model according to the configuration parameter set;

[0117] S43, using the adjusted weight configuration of each dimension, calculate the scenario-adaptive multi-dimensional deviation value between the teaching quality feature set and the historical teaching quality feature set;

[0118] S44. When the scenario-adaptive multi-dimensional deviation value is detected to converge to the preset steady-state range or a teaching stage transition is triggered, a teaching status evaluation report is generated by calling the report template based on the scenario-adaptive multi-dimensional deviation value and the current weight configuration of each dimension.

[0119] In one embodiment, after generating the teaching quality feature set, S51, the multi-dimensional score of the teaching quality feature set is calculated using the following formula:

[0120]

[0121] Where d is the dimension classification label, k is the knowledge dimension, b is the behavior dimension, c is the cognitive dimension, i and j are the interaction identifiers between different dimensions, and f is the interaction identifier between different dimensions. d For a subset of dimensional features, φ represents the elements of the inter-dimensional correlation matrix. ij ρ is the dimensional coupling strength coefficient, α is the preset coupling effect weight factor, and ρ is the multi-dimensional score of the teaching quality feature set.

[0122] In one embodiment, the method further includes:

[0123] S61, in response to the received request to obtain the teaching status assessment report, sends a teaching quality feature set update instruction to the timing processing module;

[0124] S62, based on the updated teaching quality feature set returned by the time series processing module, regenerate the teaching status assessment report.

[0125] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the functions of a big data-based dynamic evaluation system for teaching quality as described above.

[0126] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above system embodiments.

[0127] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0128] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A dynamic evaluation system for teaching quality based on big data, characterized in that, The system includes: The data acquisition module is used to acquire raw teaching time-series data streams containing student grades, classroom performance, and homework quality in real time. The time series processing module is used to add time stamps to the original teaching time series data stream and verify the timestamps to generate a teaching time stamp dataset; perform time series analysis on the teaching time stamp dataset to obtain a teaching fluctuation feature set; and perform multi-dimensional decomposition on the teaching fluctuation feature set to extract a teaching quality feature set. The real-time monitoring module is used to build a teaching monitoring model based on teaching objective parameters; Using the teaching monitoring model, the multi-dimensional deviation between the teaching quality feature set and the historical teaching quality feature set is calculated; when the multi-dimensional deviation exceeds the preset multi-dimensional joint early warning threshold, a multi-dimensional teaching early warning signal is generated.

2. The system according to claim 1, characterized in that, The timing processing module is also used for: Add a time stamp to each data unit in the original teaching time-series data stream to generate a primary time-stamped data stream; Using the primary time-stamped data stream as input, a dual-channel verification mechanism is used to check the time stamp against the physical timestamp of the data acquisition terminal, and the time-stamped dataset that passes the verification is output. Based on a preset teaching cycle template, the corresponding time-stamped datasets are categorized into teaching stages according to the time stamps of the verified time-stamped datasets, generating a teaching time-stamped dataset containing teaching stage labels.

3. The system according to claim 2, characterized in that, The timing processing module is also used for: Based on the teaching stage labels in the teaching time-stamped dataset, the teaching time-stamped dataset is grouped, and teaching time-stamped data is extracted from different groups using a preset grouping sampling rule to form time-stamped subsets; The fluctuation feature vector of teaching fluctuation is obtained by performing fluctuation feature extraction on each of the time scale subsets using the change point detection algorithm; A sensitivity analysis model is used to calculate the correlation of teaching fluctuation feature vectors between different teaching stages, and a cross-stage correlation feature matrix is ​​generated. The teaching fluctuation feature vector and the cross-stage correlation feature matrix are combined to generate the teaching fluctuation feature set.

4. The system according to claim 3, characterized in that, The timing processing module is also used for: The teaching fluctuation feature set is decomposed in parallel according to the knowledge dimension, behavior dimension and cognition dimension to generate corresponding dimensional feature subsets; Calculate the statistical correlation measure between each pair of the knowledge dimension feature subset, behavior dimension feature subset, and cognitive dimension feature subset, and generate an inter-dimensional correlation matrix; Using the generated inter-dimensional correlation matrix, a dimensional coupling strength coefficient model is constructed through a preset weighted fusion algorithm; The teaching quality feature set is generated by integrating the inter-dimensional correlation matrix and the dimensional coupling strength coefficient model.

5. The system according to claim 1, characterized in that, The real-time monitoring module is also used for: Based on a pre-set teaching scenario configuration parameter library, in response to the multi-dimensional teaching early warning signal or the preset teaching stage conversion instruction, the configuration parameter set corresponding to the current teaching scenario identifier is matched. Adjust the weight configuration of each dimension in the teaching monitoring model according to the set of configuration parameters; Using the adjusted weights of each dimension, the scenario-adaptive multidimensional deviation value between the teaching quality feature set and the historical teaching quality feature set is calculated. When the scenario-adaptive multi-dimensional deviation value is detected to converge to the preset steady-state range or a teaching stage transition is triggered, a teaching status evaluation report is generated by calling the report template based on the scenario-adaptive multi-dimensional deviation value and the current weight configuration of each dimension.

6. The system according to claim 4, characterized in that, After generating the teaching quality feature set, the real-time monitoring module calculates the multi-dimensional score of the teaching quality feature set using the following formula: Where d is the dimension classification label, k is the knowledge dimension, b is the behavior dimension, c is the cognitive dimension, i and j are the interaction identifiers between different dimensions, and f is the interaction identifier between different dimensions. d For the dimensional feature subset, φ is an element of the correlation matrix between the dimensions. ij ρ is the dimensional coupling strength coefficient, α is the preset coupling effect weight factor, and ρ is the multi-dimensional score of the teaching quality feature set.

7. The system according to claim 1, characterized in that, The real-time monitoring module is also used for: In response to the received request for obtaining the teaching status assessment report, a teaching quality feature set update instruction is sent to the time-series processing module; Based on the updated teaching quality feature set returned by the time-series processing module, a new teaching status assessment report is generated.

8. A dynamic evaluation method for teaching quality based on big data, characterized in that, The method includes: Real-time acquisition of raw teaching time-series data streams including student grades, classroom performance, and homework quality; The original teaching time-series data stream is appended with time stamps and the timestamps are verified to generate a teaching time-stamped dataset; Time series analysis was performed on the teaching time-scaled dataset to obtain a set of teaching fluctuation characteristics; The teaching fluctuation feature set is decomposed in multiple dimensions to extract the teaching quality feature set; Based on teaching objective parameters, a teaching monitoring model is constructed; Using the teaching monitoring model, calculate the multi-dimensional deviation between the teaching quality feature set and the historical teaching quality feature set; When the multi-dimensional deviation exceeds the preset multi-dimensional joint early warning threshold, a multi-dimensional teaching early warning signal is generated.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it performs the functions of the system according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it performs the functions of the system according to any one of claims 1 to 7.

Images