Intelligent teaching management system for applied mathematics

Abstract

The invention belongs to the technical field of intelligent teaching, and discloses an applied mathematics intelligent teaching management system, which comprises the steps of performing data recording on a learning process of a target student to obtain corresponding learning record data; performing data integration processing on the obtained learning record data to obtain a corresponding knowledge clustering set, and inputting the knowledge clustering set into a pre-constructed time sequence analysis model to obtain corresponding learning progress information; constructing a corresponding personalized learning scheme for the learning process of the target student based on the obtained learning progress information; according to the method, the execution process of the corresponding personalized learning scheme is supervised, feedback is performed based on the supervision result, and the personalized learning scheme can be constructed for students by comprehensively recording and analyzing the learning process of the students, so that the learning effect and the teaching quality are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent teaching technology, and more specifically, to an applied mathematics intelligent teaching management system. Background Art

[0002] In the current education field, improving teaching quality and student learning outcomes is a goal that educators and researchers are constantly pursuing. However, the traditional teaching model has exposed obvious shortcomings in many aspects and there is a large gap with the expected educational results. This provides the background for the birth of the applied mathematics intelligent teaching management system.

[0003] Compared with existing technologies, in traditional teaching, teachers’ understanding of students’ learning situation often relies on test scores and limited classroom observations. This method is difficult to fully and meticulously grasp the students’ learning process, and it is impossible to obtain multi-dimensional data such as the number of clicks on learning materials, the length of video viewing, and the time to complete exercises. Due to the lack of accurate learning situation analysis, it is difficult for teachers to formulate effective teaching strategies based on individual differences of students, resulting in uneven teaching results. In addition, most schools adopt a unified teaching schedule and teaching content, ignoring the differences in students’ knowledge mastery and learning speed. Excellent students may feel that the teaching content is too simple and cannot meet their learning needs; while students with weak foundations may not be able to keep up with the teaching pace and gradually lose interest and confidence in learning. This "one-size-fits-all" teaching model cannot achieve teaching students in accordance with their aptitude, limiting students' learning potential. At the same time, during the students' learning process, teachers cannot understand the students' learning status and difficulties in real time; when students encounter problems in learning, it is difficult to get help and guidance in time, and the accumulation of problems leads to poor learning results. Teachers also lack a comprehensive understanding of the students' learning process, making it difficult to adjust teaching strategies in a targeted manner, and unable to form an effective learning closed loop.

[0004] In view of this, the present invention proposes an applied mathematics intelligent teaching management system to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions:

[0006] An applied mathematics intelligent teaching management system, comprising:

[0007] The data collection module is used to record the learning behavior data and learning cycle data involved in the learning process of the target students and obtain the corresponding learning record data;

[0008] The data processing module annotates the corresponding learning record data with knowledge points based on the pre-acquired teaching syllabus to obtain the corresponding knowledge annotation information, constructs the corresponding data points based on the knowledge annotation information and performs information clustering processing to obtain the corresponding knowledge clustering set, which is input into the pre-constructed time series analysis module to obtain the corresponding learning progress information;

[0009] The data planning module sets a corresponding personalized learning path based on the obtained learning progress information, finds the optimal solution, obtains the corresponding optimal learning path, screens learning resources according to the marked knowledge points and classifies them into the corresponding optimal learning path to obtain the corresponding personalized learning plan;

[0010] The data feedback module is used to supervise the execution process of the corresponding personalized learning plan and provide feedback based on the supervision results.

[0011] Furthermore, the learning behavior data and learning cycle data involved in the learning process of the target student are recorded, and the process of obtaining the corresponding learning record data includes:

[0012] The data acquisition module is provided with an acquisition node and a monitoring node. The acquisition node collects the learning behavior of the target students in real time based on a preset acquisition cycle to obtain corresponding learning behavior data; the monitoring node is used to collect the learning data within the corresponding acquisition cycle to obtain corresponding learning cycle data; and the learning cycle data and learning behavior data corresponding to the corresponding acquisition cycle are counted to obtain corresponding learning record information.

[0013] Furthermore, the corresponding knowledge cluster set is obtained and input into the pre-built time series analysis module, and the process of obtaining the corresponding learning progress information includes:

[0014] Preprocessing the collected learning record information, and after the information preprocessing is completed, annotating the corresponding learning record information with knowledge points based on the knowledge point content involved in the pre-acquired teaching syllabus to obtain corresponding knowledge point annotating information;

[0015] Performing information clustering processing on the obtained knowledge point annotation information to obtain a corresponding knowledge cluster set, wherein the knowledge cluster set is composed of a plurality of knowledge cluster subsets, and the knowledge cluster subsets are composed of part of the learning record information involved in the corresponding knowledge point;

[0016] Each knowledge cluster subset in the knowledge cluster set is obtained, and is input into a pre-built time series analysis model to obtain a corresponding model output result, and based on the model output result, learning progress data corresponding to the corresponding knowledge cluster subset is obtained.

[0017] Furthermore, the obtained knowledge point annotation information is subjected to information clustering processing to obtain a corresponding knowledge clustering set, which includes:

[0018] Constructing a plurality of data points based on the knowledge point annotation information;

[0019] Obtaining the local data density and the local data distance corresponding to the corresponding data point; obtaining the quasi-clustering coefficient corresponding to each data point based on the local data density and the local data distance, and arranging them in descending order and mapping them into a two-dimensional rectangular coordinate system to obtain the corresponding cluster descending curve;

[0020] Define a data screening function to obtain all quasi-clustering centers that meet the data screening function in the corresponding data points; obtain the shortest Manhattan distance between the quasi-clustering center corresponding to the maximum quasi-clustering coefficient and other quasi-clustering centers, and compare it with the preset adaptive cutoff distance. If the quasi-clustering coefficient is less than the adaptive cutoff distance, the corresponding quasi-clustering center will be discarded; if the quasi-clustering coefficient is not less than the adaptive cutoff distance, the corresponding quasi-clustering center will be marked as a cluster point;

[0021] Get the similarity measure between the corresponding cluster points and other data points respectively In the formula, σ(x u ) and σ(x v ) represent the scaling functions of data points i and j respectively; x u and x v Respectively represent the vector representation of cluster point u and data point v in the feature space of their respective dimensions;

[0022] The obtained similarity measure is compared with a preset measure threshold. If the similarity measure is less than the measure threshold, no other operation is performed. If the similarity measure is not less than the measure threshold, the corresponding data point and the corresponding cluster point are divided into the same knowledge point subset.

[0023] Repeat the above similarity measurement acquisition process, and divide each pixel point into the corresponding cluster point based on it to obtain the corresponding knowledge cluster set.

[0024] Furthermore, the process of obtaining the local data density and the local data distance includes:

[0025] Select any data point, record it as data point i, and obtain the local data density corresponding to the corresponding data point Where, d ij represents the Manhattan distance between data point i and data point j; i, j are the indexes of the data points and i≠j; ST represents the constraint condition, d crepresents the adaptive cutoff distance; SJ represents the data set consisting of all data points;

[0026] Obtain the local data density corresponding to each data point and arrange it in descending order to obtain the corresponding density sequence;

[0027] Based on it, the local data distance corresponding to the corresponding data point i is obtained Where, d im represents the Manhattan distance between data point i and data point m; D m Represents the local data distance corresponding to data point m; m represents the descending index of the local data density in the density sequence, m≠i and m is an integer.

[0028] Furthermore, the formula for defining the data screening function is: In the formula, |k n | represents the slope of the curve between data point n and data point n+1 in the cluster descending curve; k h It represents the slope of the curve of the corresponding data point n adjacent data points; U represents the quasi-cluster center; N represents the total number of data points.

[0029] Furthermore, the construction process of the time series analysis model includes:

[0030] The backbone network of the time series analysis model is a convolutional neural network, and the basic framework of the convolutional neural network is an input layer, a convolution layer, a pooling layer and an output layer;

[0031] The input layer is used to receive input data and perform information preprocessing on it to obtain corresponding time series data;

[0032] The convolution layer is used to receive an input matrix, and perform a convolution operation on the corresponding input matrix based on a built-in convolution kernel to obtain and learn corresponding local patterns and features;

[0033] The pooling layer is used to perform pooling processing on the captured local features;

[0034] The output layer is provided with a fully connected layer for integrating the local features after the pooling process and mapping them to the pre-constructed learning progress space. Meanwhile, in combination with the obtained teaching syllabus, the corresponding learning progress trend is obtained and outputted.

[0035] The loss function of the time series analysis model is defined as In the formula, c and Z represent the index and total number of samples in the training data set, respectively, and y c and y` c Represent the actual result and expected result of the sample, wc represents the model parameters; β represents the learning efficiency factor;

[0036] Construct a training data set, divide the obtained training data set into a training set and a test set, and import the obtained training set into the constructed time series analysis model for iterative training. Use a convolutional neural network to obtain the apparent features of the corresponding time series data, and at the same time train and identify the mapping relationship between the apparent features and the learning progress. The iterative training is completed until the loss function of the time series analysis model tends to converge. After the iterative training is completed, the time series analysis model completed by the iterative training is verified based on the obtained test set. If the verification fails, continue to iteratively train it. If the verification succeeds, the model construction is completed.

[0037] Furthermore, the process of obtaining a personalized learning plan includes:

[0038] Obtain the learning progress data corresponding to each knowledge point cluster subset, and construct the learning subtasks corresponding to the corresponding knowledge point content based on it; and sort the corresponding learning subtasks according to the knowledge of the corresponding knowledge point content in the corresponding teaching outline, and obtain the corresponding task list;

[0039] Constructing a corresponding personalized learning path based on the task list, and performing optimal solution on the path to obtain a corresponding optimal learning path;

[0040] Based on the obtained optimal learning path, the task duration assigned to each learning subtask is obtained, and the corresponding learning resources are screened out from the pre-built teaching resource library in combination with the corresponding learning progress information; and they are classified into each learning subtask to obtain the corresponding personalized learning plan.

[0041] Furthermore, the process of obtaining the best learning path includes:

[0042] Constructing a corresponding initial particle population based on the personalized learning path, wherein the position of the corresponding particle in the initial particle population represents the task duration allocated when the corresponding learning subtask in the corresponding personalized learning path is executed; and the particle speed represents the adjustment direction and amplitude of the corresponding personalized learning path;

[0043] Define the fitness function In the formula, score 1 and score 0 They represent the expected test scores and real-time test scores of the target students, respectively; maxscore represents the total test score; T use and T maxThey represent the expected task duration and the maximum allowed task duration of the corresponding learning subtask respectively; α represents the weight coefficient, α>0.5, that is, in the actual application process, the importance of improving test scores is relatively higher;

[0044] Then, the fitness value of each particle in the initial particle population is obtained, and the initial position and initial velocity of each particle are iteratively updated based on the fitness value;

[0045] Repeat the above-mentioned updating process of the initial position and initial velocity until the iteration stop condition is met and the corresponding optimal learning path is obtained;

[0046] Among them, the iterative update formula of the initial velocity is V ab (t+1)=ω×V ab (t)+c1×r 1 (t)(P ab -X ab (t))+c2×r 2 (t)(G b -X ab (t));

[0047] The iterative update formula for the initial position is: ab (t+1)=X ab (t)+V ab (t+1);

[0048] Where V ab (t+1) and V ab (t) respectively represent the initial velocity of particle a before and after the tth iteration update in the b-dimensional space; X ab (t) and X ab (t+1) represent the initial positions of particle a before and after the tth iteration in the b-dimensional space; ω represents the relationship weight; c1 and c2 are learning factors, usually called acceleration constants, c1 represents the degree to which the particle learns from its own historical optimal position, and c2 represents the degree to which the particle learns from the global optimal position; P ab represents the historical optimal position of the particle; G b represents the global optimal position, r 1 (t) and r 2 (t) represents a random number between [0,1].

[0049] Furthermore, the process of supervising the execution process of the corresponding personalized learning plan and providing feedback based on the supervision results includes:

[0050] The obtained personalized learning plan is fed back to the target students, and the target students can learn the knowledge points based on the task duration and learning resources provided in the personalized learning plan; at the same time, the learning behavior data of the corresponding knowledge learning process is collected in real time based on the collection node, and the learning behavior data of the target students is analyzed based on the principal component analysis method to obtain abnormal patterns in the target students' learning process, and provide difficulty feedback based on them.

[0051] The technical effects and advantages of the intelligent teaching management system for applied mathematics of the present invention are as follows:

[0052] 1. Based on accurate learning progress information, build personalized learning plans for students. First, build and sort learning subtasks, then use the particle swarm algorithm to solve the best learning path, and then select suitable resources from the teaching resource library; match basic, general or advanced resources according to the students' mastery of knowledge points, implement teaching in accordance with their aptitude, and improve students' learning efficiency and enthusiasm.

[0053] 2. Through the data feedback module, when students study according to the personalized learning plan, the collection nodes are used to collect learning behavior data in real time, and the principal component analysis method is used to monitor anomalies; and feedback is given to teachers in a timely manner. Teachers can understand the students' learning progress, point out the advantages and disadvantages, and give suggestions for improvement, forming a learning closed loop and ensuring learning effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A schematic diagram of an applied mathematics intelligent teaching management system of the present invention;

[0055] Figure 2 It is a schematic diagram of an applied mathematics intelligent teaching management method of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0057] Example 1

[0058] See also Figure 1 As shown, the applied mathematics intelligent teaching management system described in this embodiment includes:

[0059] The data collection module is used to record the learning behavior data and learning cycle data involved in the learning process of the target students and obtain the corresponding learning record data;

[0060] The data processing module annotates the corresponding learning record data with knowledge points based on the pre-acquired teaching syllabus to obtain the corresponding knowledge annotation information, constructs the corresponding data points based on the knowledge annotation information and performs information clustering processing to obtain the corresponding knowledge clustering set, which is input into the pre-constructed time series analysis module to obtain the corresponding learning progress information;

[0061] The data planning module sets a corresponding personalized learning path based on the obtained learning progress information, finds the optimal solution, obtains the corresponding optimal learning path, screens learning resources according to the marked knowledge points and classifies them into the corresponding optimal learning path to obtain the corresponding personalized learning plan;

[0062] The data feedback module is used to supervise the execution process of the corresponding personalized learning plan and provide feedback based on the supervision results;

[0063] The modules are connected to each other via wired and / or wireless means to achieve data transmission between modules.

[0064] It should be further explained that, in the specific implementation process, the process of recording the learning process of the target students and obtaining the corresponding learning record data includes:

[0065] The data collection module is provided with a collection node and a monitoring node. The collection node collects the learning behavior of the target student in real time based on a preset collection cycle to obtain corresponding learning behavior data; the learning behavior data includes the number of times the student clicks on the learning material, the length of time the video is watched, the time spent on completing a certain exercise, etc.;

[0066] The monitoring node is used to collect learning data within a corresponding collection period to obtain corresponding learning period data, wherein the learning period data includes homework scores, test scores, completion time, etc.;

[0067] The learning cycle data and learning behavior data corresponding to the corresponding collection period are counted to obtain the corresponding learning record information.

[0068] It should be further explained that, in the specific implementation process, the acquired learning record data is subjected to data integration processing to obtain the corresponding knowledge clustering set, and the set is input into the pre-built time series analysis model to obtain the corresponding learning progress information. The process includes:

[0069] Performing information preprocessing on the collected learning record information, wherein the information preprocessing includes data cleaning and data standardization processing, wherein the data cleaning refers to removing duplicate data in the corresponding learning record information; and the standardization processing refers to normalizing the corresponding learning record information to make different types of data comparable;

[0070] After the information preprocessing is completed, the corresponding learning record information is annotated with knowledge points based on the knowledge point content involved in the pre-acquired teaching syllabus to obtain corresponding knowledge point annotating information, wherein the knowledge point annotating information refers to text annotation of the corresponding learning record information based on the knowledge point content involved in the teaching syllabus. For example, when the annotated learning record information records the number of times the student clicks on the learning material, the knowledge point content involved in the learning material is clarified based on the teaching syllabus and text annotation is performed;

[0071] The obtained knowledge point annotation information is clustered to obtain a corresponding knowledge cluster set; wherein the process of obtaining the knowledge cluster set includes:

[0072] Constructing a plurality of data points based on the knowledge point annotation information, wherein the data point is at least one of the learning record information such as the test score, the homework score, the time spent on completing a certain exercise after the knowledge point annotation is completed;

[0073] It should be further explained that, in the specific implementation process, taking the test score as an example, there are many knowledge points involved in the corresponding test process, so when the corresponding test score is annotated with knowledge points, it is first divided according to the test question type to obtain the corresponding test question type score, and the knowledge point content is annotated (that is, the test score is divided into the test question type score under each knowledge point content), and the corresponding data point is constructed based on it;

[0074] Select any data point, record it as data point i, and obtain the local data density corresponding to the corresponding data point Where, d ij represents the Manhattan distance between data point i and data point j, which can be used to represent the sum of the distances between two data points in various dimensions; i, j are both indexes of the data points and i≠j; ST represents the constraint condition, d c Represents adaptive cutoff distance; the cutoff distance is used to define the neighborhood relationship between data points, representing a distance range defined in the data space with a certain data point as the center; it is pre-set by personnel in this field; SJ represents a data set consisting of all data points;

[0075] Get the local data density corresponding to each data point and arrange it in descending order to obtain the corresponding density sequence in, N represents the total number of data points;

[0076] Then, based on it, the local data distance corresponding to the corresponding data point i is obtained Where, d imrepresents the Manhattan distance between data point i and data point m; D m Represents the local data distance corresponding to data point m; m represents the descending index of the local data density in the density sequence, m≠i and m is an integer;

[0077] Then, based on the local data density and the local data distance, the quasi-clustering coefficient corresponding to each data point is obtained, and the quasi-clustering coefficient is arranged in descending order and mapped into a two-dimensional rectangular coordinate system to obtain a corresponding cluster descending curve;

[0078] Define a data filtering function: In the formula, k n | represents the slope of the curve between data point n and data point n+1 in the cluster descending curve; k h represents the slope of the curve of the adjacent data points of the corresponding data point n; U represents the quasi-cluster center; wherein the definition expressed by the above data screening function is that if the slope of the curve at the data point n satisfies Then the corresponding data point n is output as the quasi-clustering center;

[0079] Obtain all quasi-clustering centers that satisfy the data screening function in the corresponding data points; obtain the shortest Manhattan distance between the quasi-clustering center corresponding to the maximum quasi-clustering coefficient and other quasi-clustering centers, and compare it with the adaptive cutoff distance. If the quasi-clustering coefficient is less than the adaptive cutoff distance, the corresponding quasi-clustering center is discarded; if the quasi-clustering coefficient is not less than the adaptive cutoff distance, the corresponding quasi-clustering center is marked as a cluster point;

[0080] Then, we obtain the similarity measure between the corresponding cluster points and other data points respectively. In the formula, σ(x u ) and σ(x v ) represent the scale functions of data points i and j respectively, which are used to control the size of the influence domain of the clustering point. The scale function is set in advance according to actual needs; x u and x v Respectively represent the vector representation of cluster point u and data point v in the feature space of their respective dimensions;

[0081] Then, the obtained similarity measure is compared with a preset measure threshold. If the similarity measure is less than the measure threshold, no other operation is performed. If the similarity measure is not less than the measure threshold, the corresponding data point and the corresponding cluster point are divided into the same knowledge point subset.

[0082] Repeat the above similarity measurement acquisition process, and divide each pixel point into corresponding cluster points based on it, to obtain a corresponding knowledge cluster set, wherein the knowledge cluster set is composed of several knowledge cluster subsets, and the knowledge cluster subset is composed of part of the learning record information involved in the corresponding knowledge point;

[0083] Further, each knowledge cluster subset in the knowledge cluster set is obtained, and is input into a pre-built time series analysis model to obtain a corresponding model output result, and based on the model output result, learning progress data corresponding to the corresponding knowledge cluster subset is obtained, wherein the learning progress data includes the learning progress of the target student in the corresponding knowledge point content and the mastering speed and degree of the corresponding knowledge point;

[0084] The construction process of the time series analysis model includes:

[0085] The backbone network of the time series analysis model is a convolutional neural network, and the basic framework of the convolutional neural network is an input layer, a convolution layer, a pooling layer and an output layer;

[0086] The input layer is used to receive input data and perform information preprocessing on it to obtain corresponding time series data; the information preprocessing refers to sorting the data parameters in the input data according to the time sequence, and constructing a corresponding two-dimensional matrix based on it. At the same time, the two-dimensional matrices constructed by data parameters of different data types are matrix spliced ​​to obtain the corresponding input matrix;

[0087] The convolution layer is used to receive an input matrix and perform a convolution operation on the corresponding input matrix based on a built-in convolution kernel to capture the relationship between adjacent time steps, obtain corresponding local patterns and features, and learn;

[0088] The pooling layer is used to perform pooling processing on the captured local features to retain important features and reduce the computational complexity and overfitting risk of the model;

[0089] The output layer is provided with a fully connected layer for integrating the local features after the pooling process and mapping them to the pre-constructed learning progress space. Meanwhile, in combination with the obtained teaching syllabus, the corresponding learning progress trend is obtained and outputted.

[0090] The loss function of the time series analysis model is defined as In the formula, c and Z represent the index and total number of samples in the training data set, respectively, and y c and y` c Represent the actual result and expected result of the sample, w c represents the model parameters; β represents the learning efficiency factor;

[0091] Obtaining historical learning record information corresponding to a number of different students, and processing each historical learning record information based on the above information clustering process to obtain a corresponding historical knowledge point cluster set, and manually annotating the knowledge point cluster subset corresponding to each knowledge point content in the corresponding historical knowledge point cluster set, wherein the manual annotation refers to manually reviewing the data in the corresponding knowledge cluster subset, and annotating the learning progress information based on the review result;

[0092] Furthermore, a corresponding training data set is constructed based on the manually annotated knowledge point clustering set, the obtained training data set is divided into a training set and a test set, and the obtained training set is imported into the constructed time series analysis model for iterative training. The apparent features of the corresponding time series data are obtained through a convolutional neural network, and the mapping relationship between the apparent features and the learning progress is trained and recognized at the same time; the iterative training is completed until the loss function of the time series analysis model tends to converge, and after the iterative training is completed, the time series analysis model completed by the iterative training is verified based on the obtained test set. If the verification fails, the iterative training continues. If the verification succeeds, the model construction is completed.

[0093] It should be further explained that, in the specific implementation process, the process of obtaining a personalized learning plan includes:

[0094] Obtain the learning progress data corresponding to each knowledge point cluster subset, and construct the learning subtasks corresponding to the corresponding knowledge point content based on it; and sort the corresponding learning subtasks according to the knowledge of the corresponding knowledge point content in the corresponding teaching outline, and obtain the corresponding task list;

[0095] Then, a corresponding personalized learning path is constructed based on the task list, and an optimal solution is performed on the personalized learning path to obtain a corresponding optimal learning path;

[0096] The process of obtaining the best learning path includes:

[0097] Constructing a corresponding initial particle population based on the personalized learning path, wherein the position of the corresponding particle in the initial particle population represents the task duration allocated when the corresponding learning subtask in the corresponding personalized learning path is executed; and the particle speed represents the adjustment direction and amplitude of the corresponding personalized learning path;

[0098] Define the fitness function In the formula, score 1 and score 0 They represent the expected test scores and real-time test scores of the target students, respectively; maxscore represents the total test score; T use and T maxThey represent the expected task duration and the maximum allowed task duration of the corresponding learning subtask respectively; α represents the weight coefficient, α>0.5, that is, in the actual application process, the importance of improving test scores is relatively higher;

[0099] Then, the fitness value of each particle in the initial particle population is obtained, and the initial position and initial velocity of each particle are iteratively updated based on the fitness value;

[0100] Repeat the above-mentioned updating process of the initial position and initial velocity until the iteration stopping condition is met to obtain the corresponding optimal learning path; the iteration stopping condition refers to reaching a preset maximum number of iterations or the particle swarm gradually converges to the desired target;

[0101] Among them, the iterative update formula of the initial velocity is V ab (t+1)=ω×V ab (t)+c1×r 1 (t)(P ab -X ab (t))+c2×r 2 (t)(G b -X ab (t));

[0102] The iterative update formula for the initial position is: ab (t+1)=X ab (t)+V ab (t+1);

[0103] Where V ab (t+1) and V ab (t) respectively represent the initial velocity of particle a before and after the tth iteration update in the b-dimensional space; X ab (t) and X ab (t+1) represent the initial positions of particle a before and after the tth iteration in the b-dimensional space; ω represents the relationship weight; c1 and c2 are learning factors, usually called acceleration constants, c1 represents the degree to which the particle learns from its own historical optimal position, and c2 represents the degree to which the particle learns from the global optimal position; P ab represents the historical optimal position of the particle; G b represents the global optimal position, r 1 (t) and r 2 (t) represents a random number between [0,1];

[0104] Then, based on the obtained optimal learning path, the task duration assigned to each learning subtask is obtained, and the corresponding learning resources are screened out from the pre-built teaching resource library in combination with the corresponding learning progress information. The learning resources include exercises, teaching videos, and teaching materials, etc.; and the resources are classified into each learning subtask to obtain a corresponding personalized learning plan;

[0105] In one embodiment of the present invention, for example, when the learning subtask is a corresponding algebra application, learning resources of different knowledge levels are selected based on the students' mastery of knowledge points, and learning planning is performed based on the assigned task duration. The knowledge levels include three stages: basic, general, and advanced. The higher the mastery of the corresponding knowledge points, the higher the knowledge level corresponding to the corresponding learning resources.

[0106] It should be further explained that, in the specific implementation process, the construction process of the teaching resource library includes:

[0107] Resource nodes and distributed databases are set up. Teachers can upload and store various teaching resources they own in the corresponding distributed database according to the specified format based on the resource nodes to obtain the corresponding teaching resource library. The teaching resource library will classify the received teaching resources according to the knowledge points, teaching outlines, resource types, etc. involved, so as to facilitate subsequent access by students and teachers.

[0108] It should be further explained that, in the specific implementation process, the process of supervising the execution of the corresponding personalized learning plan and providing feedback based on the supervision results includes:

[0109] The obtained personalized learning plan is fed back to the target students, and the target students can learn the knowledge points based on the task duration and learning resources provided in the personalized learning plan; at the same time, the learning behavior data of the corresponding knowledge learning process is collected in real time based on the collection node, and the learning behavior data of the target students is analyzed based on the principal component analysis method to obtain abnormal patterns in the target students' learning process, and difficulty feedback is given based on them; for example: when a student stays at a certain time node in a teaching video or a certain exercise for a long time, repeats incorrect operations many times, etc., it is determined that the student encounters difficulties; feedback is given to the teacher in charge of the target student, and the corresponding teacher intervenes in the learning of the knowledge point, and the current learning progress information of the target student is fed back to the teacher in charge, so as to point out the strengths and weaknesses of the target student and give suggestions for improvement.

[0110] The present invention uses a data collection module to collect rich data from two dimensions: learning behavior and learning cycle, such as the number of times learning materials are clicked, homework scores, etc., through collection nodes and monitoring nodes. After data cleaning, standardization processing and knowledge point labeling, data points are constructed and complex clustering analysis is performed. Combined with the time series analysis model, the learning progress, mastering speed and degree of students at each knowledge point can be accurately obtained, providing a scientific basis for teaching decisions, and constructing personalized learning plans for them, thereby improving learning effects and teaching quality.

[0111] Example 2

[0112] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides an applied mathematics intelligent teaching management method, which is characterized by comprising:

[0113] Step 1: Record the learning behavior data and learning cycle data involved in the learning process of the target students to obtain corresponding learning record data;

[0114] Step 2: Based on the pre-acquired teaching syllabus, the corresponding learning record data is labeled with knowledge points to obtain the corresponding knowledge labeling information, based on which the corresponding data points are constructed and information clustering is performed to obtain the corresponding knowledge clustering set, which is input into the pre-built time series analysis module to obtain the corresponding learning progress information;

[0115] Step 3: Set a corresponding personalized learning path based on the obtained learning progress information, find the optimal solution, obtain the corresponding optimal learning path, screen the learning resources according to the marked knowledge points and classify them into the corresponding optimal learning path to obtain the corresponding personalized learning plan;

[0116] Step 4: Monitor the implementation process of the corresponding personalized learning plan and provide feedback based on the supervision results.

[0117] Example 3

[0118] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the applied mathematics intelligent teaching management system provided above is implemented.

[0119] Since the electronic device introduced in this embodiment is an electronic device used to implement an applied mathematics intelligent teaching management system in the embodiment of this application, based on the applied mathematics intelligent teaching management system introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as the technical personnel of this field implement the electronic device used in the applied mathematics intelligent teaching management system in the embodiment of this application, it belongs to the scope of protection of this application.

[0120] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0121] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. An applied mathematics intelligent teaching management system, characterized in that: include: The data collection module is used to record the learning behavior data and learning cycle data involved in the learning process of the target students and obtain the corresponding learning record data; The data processing module annotates the corresponding learning record data with knowledge points based on the pre-acquired teaching syllabus to obtain the corresponding knowledge annotation information, constructs the corresponding data points based on the knowledge annotation information and performs information clustering processing to obtain the corresponding knowledge clustering set, which is input into the pre-constructed time series analysis module to obtain the corresponding learning progress information; The data planning module sets a corresponding personalized learning path based on the obtained learning progress information, finds the optimal solution, obtains the corresponding optimal learning path, screens learning resources according to the marked knowledge points and classifies them into the corresponding optimal learning path to obtain the corresponding personalized learning plan; The data feedback module is used to supervise the execution process of the corresponding personalized learning plan and provide feedback based on the supervision results.

2. The intelligent teaching management system for applied mathematics according to claim 1 is characterized in that: The process of recording the learning behavior data and learning cycle data involved in the learning process of the target students and obtaining the corresponding learning record data includes: The data acquisition module is provided with an acquisition node and a monitoring node. The acquisition node collects the learning behavior of the target students in real time based on a preset acquisition cycle to obtain corresponding learning behavior data; the monitoring node is used to collect the learning data within the corresponding acquisition cycle to obtain corresponding learning cycle data; and the learning cycle data and learning behavior data corresponding to the corresponding acquisition cycle are counted to obtain corresponding learning record information.

3. The intelligent teaching management system for applied mathematics according to claim 2 is characterized in that: The process of obtaining the corresponding knowledge cluster set and inputting it into the pre-built time series analysis module to obtain the corresponding learning progress information includes: Preprocessing the collected learning record information, and after the information preprocessing is completed, annotating the corresponding learning record information with knowledge points based on the knowledge point content involved in the pre-acquired teaching syllabus to obtain corresponding knowledge point annotating information; Performing information clustering processing on the obtained knowledge point annotation information to obtain a corresponding knowledge cluster set, wherein the knowledge cluster set is composed of a plurality of knowledge cluster subsets, and the knowledge cluster subsets are composed of part of the learning record information involved in the corresponding knowledge point; Each knowledge cluster subset in the knowledge cluster set is obtained, and is input into a pre-built time series analysis model to obtain a corresponding model output result, and based on the model output result, learning progress data corresponding to the corresponding knowledge cluster subset is obtained.

4. The intelligent teaching management system for applied mathematics according to claim 3 is characterized in that: The process of performing information clustering processing on the obtained knowledge point annotation information to obtain the corresponding knowledge clustering set includes: Constructing a plurality of data points based on the knowledge point annotation information; Obtaining the local data density and the local data distance corresponding to the corresponding data point; obtaining the quasi-clustering coefficient corresponding to each data point based on the local data density and the local data distance, and arranging them in descending order and mapping them into a two-dimensional rectangular coordinate system to obtain the corresponding cluster descending curve; Define a data screening function to obtain all quasi-clustering centers that meet the data screening function in the corresponding data points; obtain the shortest Manhattan distance between the quasi-clustering center corresponding to the maximum quasi-clustering coefficient and other quasi-clustering centers, and compare it with the preset adaptive cutoff distance. If the quasi-clustering coefficient is less than the adaptive cutoff distance, the corresponding quasi-clustering center will be discarded; if the quasi-clustering coefficient is not less than the adaptive cutoff distance, the corresponding quasi-clustering center will be marked as a cluster point; Get the similarity measure between the corresponding cluster points and other data points respectively In the formula, σ(x u ) and σ(x v ) represent the scaling functions of data points i and j respectively; x u and x v Respectively represent the vector representation of cluster point u and data point v in the feature space of their respective dimensions; The obtained similarity measure is compared with a preset measure threshold. If the similarity measure is less than the measure threshold, no other operation is performed. If the similarity measure is not less than the measure threshold, the corresponding data point and the corresponding cluster point are divided into the same knowledge point subset. Repeat the above similarity measurement acquisition process, and divide each pixel point into the corresponding cluster point based on it to obtain the corresponding knowledge cluster set.

5. The intelligent teaching management system for applied mathematics according to claim 4 is characterized in that: The process of obtaining local data density and local data distance includes: Select any data point, record it as data point i, and obtain the local data density corresponding to the corresponding data point Where, d ij represents the Manhattan distance between data point i and data point j; i, j are the indexes of the data points and i≠j; ST represents the constraint condition, d c represents the adaptive cutoff distance; SJ represents the data set consisting of all data points; Obtain the local data density corresponding to each data point and arrange it in descending order to obtain the corresponding density sequence; Based on it, the local data distance D corresponding to the corresponding data point i is obtained i = Where, d im represents the Manhattan distance between data point i and data point m; D m Represents the local data distance corresponding to data point m; m represents the descending index of the local data density in the density sequence, m≠i and m is an integer.

6. The intelligent teaching management system for applied mathematics according to claim 5, characterized in that: The formula for defining the data filtering function is: In the formula, k n | represents the slope of the curve between data point n and data point n+1 in the cluster descending curve; k h It represents the slope of the curve of the corresponding data point n adjacent data points; U represents the quasi-cluster center; N represents the total number of data points.

7. The intelligent teaching management system for applied mathematics according to claim 6 is characterized in that: The construction process of the time series analysis model includes: The backbone network of the time series analysis model is a convolutional neural network, and the basic framework of the convolutional neural network is an input layer, a convolution layer, a pooling layer and an output layer; The input layer is used to receive input data and perform information preprocessing on it to obtain corresponding time series data; The convolution layer is used to receive an input matrix, and perform a convolution operation on the corresponding input matrix based on a built-in convolution kernel to obtain and learn corresponding local patterns and features; The pooling layer is used to perform pooling processing on the captured local features; The output layer is provided with a fully connected layer for integrating the local features after the pooling process and mapping them to the pre-constructed learning progress space. Meanwhile, in combination with the obtained teaching syllabus, the corresponding learning progress trend is obtained and outputted. The loss function of the time series analysis model is defined as In the formula, c and Z represent the index and total number of samples in the training data set, respectively, and y c and y` c Represent the actual result and expected result of the sample, w c represents the model parameters; β represents the learning efficiency factor; Construct a training data set, divide the obtained training data set into a training set and a test set, and import the obtained training set into the constructed time series analysis model for iterative training. Use a convolutional neural network to obtain the apparent features of the corresponding time series data, and at the same time train and identify the mapping relationship between the apparent features and the learning progress. The iterative training is completed until the loss function of the time series analysis model tends to converge. After the iterative training is completed, the time series analysis model completed by the iterative training is verified based on the obtained test set. If the verification fails, continue to iteratively train it. If the verification succeeds, the model construction is completed.

8. The intelligent teaching management system for applied mathematics according to claim 7 is characterized in that: The process of obtaining a personalized learning plan includes: Obtain the learning progress data corresponding to each knowledge point cluster subset, and construct the learning subtasks corresponding to the corresponding knowledge point content based on it; and sort the corresponding learning subtasks according to the knowledge of the corresponding knowledge point content in the corresponding teaching outline, and obtain the corresponding task list; Constructing a corresponding personalized learning path based on the task list, and performing optimal solution on the path to obtain a corresponding optimal learning path; Based on the obtained optimal learning path, the task duration assigned to each learning subtask is obtained, and the corresponding learning resources are screened out from the pre-built teaching resource library in combination with the corresponding learning progress information; and they are classified into each learning subtask to obtain the corresponding personalized learning plan.

9. The intelligent teaching management system for applied mathematics according to claim 8, characterized in that: The process of obtaining the best learning path includes: Constructing a corresponding initial particle population based on the personalized learning path, wherein the position of the corresponding particle in the initial particle population represents the task duration allocated when the corresponding learning subtask in the corresponding personalized learning path is executed; and the particle speed represents the adjustment direction and amplitude of the corresponding personalized learning path; Define the fitness function In the formula, score1 and score0 represent the expected test score and real-time test score of the target student respectively, and maxscore represents the total test score; T use and T max They represent the expected task duration and the maximum allowed task duration of the corresponding learning subtask respectively; α represents the weight coefficient, α>0.5, that is, in the actual application process, the importance of improving test scores is relatively higher; Then, the fitness value of each particle in the initial particle population is obtained, and the initial position and initial velocity of each particle are iteratively updated based on the fitness value; Repeat the above-mentioned updating process of the initial position and initial velocity until the iteration stop condition is met and the corresponding optimal learning path is obtained; Among them, the iterative update formula of the initial velocity is V ab (t+1)=ω×V ab (t)+c1×r1(t)(P ab -X ab (t))+c2×r2(t)(G b -X ab (t)); The iterative update formula for the initial position is: ab (t+1)=X ab (t)+V ab (t+1); Where V ab (t+1) and V ab (t) respectively represent the initial velocity of particle a before and after the tth iteration update in the b-dimensional space; X ab (t) and X ab (t+1) represent the initial positions of particle a before and after the tth iteration in the b-dimensional space; ω represents the relationship weight; c1 and c2 are learning factors, usually called acceleration constants, c1 represents the degree to which the particle learns from its own historical optimal position, and c2 represents the degree to which the particle learns from the global optimal position; P ab represents the historical optimal position of the particle; G b represents the global optimal position, r1(t) and r2(t) represent random numbers between [0,1].

10. The intelligent teaching management system for applied mathematics according to claim 9, characterized in that: The process of supervising the implementation of the corresponding personalized learning plan and providing feedback based on the supervision results includes: The obtained personalized learning plan is fed back to the target students, and the target students can learn the knowledge points based on the task duration and learning resources provided in the personalized learning plan; at the same time, the learning behavior data of the corresponding knowledge learning process is collected in real time based on the collection node, and the learning behavior data of the target students is analyzed based on the principal component analysis method to obtain abnormal patterns in the target students' learning process, and provide difficulty feedback based on them.

Images