Group teaching strategy generation method based on smart learning cloud platform

Abstract

The invention is suitable for the technical field of group teaching, and particularly relates to a group teaching strategy generation method based on a smart learning cloud platform, and the method comprises the steps: obtaining the teaching behavior data, biological data and teaching progress data of a plurality of classes through the smart learning cloud platform; according to the teaching progress data of the plurality of classes, carrying out space-time alignment processing on the teaching stages of the plurality of classes to obtain a teaching stage sequence; based on each teaching stage of the teaching stage sequence, performing group anomaly detection on a plurality of classes to obtain an anomaly detection result; and generating a group teaching strategy based on the teaching behavior data and the biological data of the teaching stage in the anomaly detection result. According to the method, group problems can be found and corrected in time, and teachers are helped to make more accurate teaching decisions, so that the overall teaching quality and learning effect are improved.

Description

Technical Field

[0001] This application belongs to the technical field of group teaching, and particularly relates to a method for generating group teaching strategies based on a smart learning cloud platform. Background Art

[0002] With the development of technology, online education platforms and network education resources are increasing day by day. Teachers can understand the learning situation of students through education platforms and implement group teaching models through education platforms. This model connects teachers and students through the platform. Teachers can teach multiple students at the same time, and students can participate in interactive learning through the platform. The platform usually provides functions such as video live broadcast, courseware sharing, and discussion areas to promote communication and collaboration between teachers and students, and among students.

[0003] In the prior art, in teaching environments with multiple classes or across campuses, the backgrounds and learning progress of students vary. Although group teaching based on education platforms enables teachers to access more data, how to systematically identify common problems between different classes and campuses remains a challenge. Although education platforms have achieved online interaction and courseware sharing, teachers still need to spend a lot of time analyzing the learning situation of students to design teaching plans. Even though time can be saved by sampling and checking students' homework or test scores, it is easy for teachers to overlook some hidden misunderstandings and confusion of knowledge points.

[0004] In summary, when using education platforms to conduct group teaching for students, there is a problem that the teaching quality is reduced due to the lag in discovering group learning blind spots. Summary of the Invention

[0005] The embodiments of this application provide a method for generating group teaching strategies based on a smart learning cloud platform, which can solve the problem in the related art that when using education platforms to conduct group teaching for students, the teaching quality is reduced due to the lag in discovering group learning blind spots.

[0006] In a first aspect, the embodiments of this application provide a method for generating group teaching strategies based on a smart learning cloud platform, including: Obtaining teaching behavior data, biological data, and teaching progress data of multiple classes through a smart learning cloud platform; wherein, the teaching progress data includes taught knowledge points and corresponding teaching dates; Performing spatio-temporal alignment processing on the teaching stages of multiple classes according to the teaching progress data of multiple classes to obtain a teaching stage sequence; wherein, the teaching stage sequence includes multiple teaching stages, and each teaching stage includes teaching behavior data and biological data of multiple classes in the same teaching stage; For each teaching stage in the teaching stage sequence, perform group anomaly detection on multiple classes to obtain anomaly detection results; wherein, the anomaly detection results include the teaching stages with group anomalies. Generate a group teaching strategy based on the teaching behavior data and biological data of the teaching stages in the anomaly detection results.

[0007] The above technical solutions in the embodiments of the present application have at least the following technical effects: The method for generating a group teaching strategy based on a smart learning cloud platform provided by the present application first obtains the teaching behavior data, biological data, and teaching progress data (taught knowledge points and corresponding teaching dates) of multiple classes through the smart learning cloud platform, and then performs spatio-temporal alignment processing on the teaching stages of multiple classes according to the teaching progress data of multiple classes to obtain a teaching stage sequence (the teaching stage sequence includes multiple teaching stages, and each teaching stage includes the teaching behavior data and biological data of multiple classes in the same teaching stage). Then, for each teaching stage in the teaching stage sequence, perform group anomaly detection on multiple classes to obtain anomaly detection results (teaching stages with group anomalies). Finally, generate a group teaching strategy based on the teaching behavior data and biological data of the teaching stages in the anomaly detection results. This method can timely identify the teaching stages with group learning blind spots or learning obstacles, and then adjust the teaching strategy, which is beneficial to timely discovering problems and correcting them, and avoiding the decline of teaching quality. This method can not only improve the learning progress and comprehension ability of individual students, but also optimize the learning atmosphere and interaction of the entire class. This optimization targets collective learning blind spots and difficulties, avoiding some students being ignored in group learning, and promoting the learning progress of all students. The generated group teaching strategy can better meet the different learning needs of different classes and students in the same teaching stage, help teachers make more accurate teaching decisions, and thus improve the overall teaching quality and learning effect.

[0008] In a second aspect, an embodiment of the present application provides a device for generating a group teaching strategy based on a smart learning cloud platform, including: An acquisition unit, configured to obtain the teaching behavior data, biological data, and teaching progress data of multiple classes through the smart learning cloud platform; wherein, the teaching progress data includes the taught knowledge points and the corresponding teaching dates. A spatio-temporal alignment unit, configured to perform spatio-temporal alignment processing on the teaching stages of multiple classes according to the teaching progress data of multiple classes to obtain a teaching stage sequence; wherein, the teaching stage sequence includes multiple teaching stages, and each teaching stage includes the teaching behavior data and biological data of multiple classes in the same teaching stage. A group anomaly detection unit, configured to perform group anomaly detection on multiple classes based on each teaching stage of the teaching stage sequence, and obtain an anomaly detection result; wherein, the anomaly detection result includes the teaching stages with group anomalies. A group teaching strategy generation unit, configured to generate a group teaching strategy based on the teaching behavior data and biological data of the teaching stages in the anomaly detection result.

[0009] In a third aspect, an embodiment of the present application provides 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 method described in any one of the embodiments in the first aspect is implemented.

[0010] It can be understood that the beneficial effects of the above second aspect to the third aspect can refer to the relevant descriptions in the first aspect above, and will not be elaborated here. Description of the Drawings

[0011] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0012] Figure 1 It is a schematic flowchart of a method for generating a group teaching strategy based on a smart learning cloud platform provided by an embodiment of the present application; Figure 2 It is a schematic implementation flowchart of group anomaly detection in the method for generating a group teaching strategy based on a smart learning cloud platform provided by an embodiment of the present application; Figure 3 It is a schematic overall architecture diagram of the method for generating a group teaching strategy based on a smart learning cloud platform provided by an embodiment of the present application. Detailed Embodiments

[0013] In the following description, for the purpose of illustration rather than limitation, specific details such as specific system architectures and technologies are proposed to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, the detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application.

[0014] It should be understood that, as used in the specification of this application and the appended claims, the term "comprising" indicates the presence of the described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or their groups.

[0015] It should also be understood that the term "and / or" as used in the specification of this application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0016] In the related art, in a teaching environment with multiple classes or across campuses, the backgrounds and learning progress of students vary. Although group teaching based on an educational platform enables teachers to access more data, how to systematically identify common problems among different classes and campuses remains a challenge. Especially when there is a lack of a unified data analysis and management tool, it is difficult for teachers to conduct a comprehensive analysis of students' learning conditions through a unified platform, and thus it is difficult to discover the general problems of students in different regions. Although the educational platform has realized online interaction and courseware sharing, teachers still need to spend a lot of time analyzing students' learning situations to design teaching plans. Even though time can be saved by sampling and checking students' homework or test scores, it is easy for teachers to overlook some hidden misunderstandings and confusion of knowledge points. For example, although some students answer questions correctly, they may have conceptual confusion or insufficient understanding of some details, and these problems are often difficult to be discovered and corrected in a timely manner when only through sampling checks.

[0017] To solve the above problems, the embodiments of the present application provide a method for generating a group teaching strategy based on a smart learning cloud platform. In this method, first, teaching behavior data, biological data, and teaching progress data (taught knowledge points and corresponding teaching dates) of multiple classes are obtained through the smart learning cloud platform. Then, according to the teaching progress data of multiple classes, space-time alignment processing is performed on the teaching stages of multiple classes to obtain a teaching stage sequence (the teaching stage sequence includes multiple teaching stages, and each teaching stage includes the teaching behavior data and biological data of multiple classes in the same teaching stage). Next, based on each teaching stage of the teaching stage sequence, group anomaly detection is performed on multiple classes to obtain an anomaly detection result (the teaching stage with group anomalies). Finally, based on the teaching behavior data and biological data of the teaching stages in the anomaly detection result, a group teaching strategy is generated. This method can timely identify the teaching stages with group learning blind spots or learning obstacles, and then adjust the teaching strategy, which is beneficial to timely discovering problems and correcting them, and avoiding the decline of teaching quality. This method can not only improve the learning progress and comprehension ability of individual students, but also optimize the learning atmosphere and interaction of the entire class. This optimization targets collective learning blind spots and difficulties, avoiding some students being ignored in group learning, and promoting the learning progress of all students. The generated group teaching strategy can better meet the different learning needs of different classes and students in the same teaching stage, help teachers make more accurate teaching decisions, and thus improve the overall teaching quality and learning effect.

[0018] The method for generating a group teaching strategy based on a smart learning cloud platform provided by the embodiments of the present application can be applied to an electronic device. At this time, the electronic device is the execution subject of the method for generating a group teaching strategy based on a smart learning cloud platform provided by the embodiments of the present application. The embodiments of the present application do not impose any restrictions on the specific type of the electronic device.

[0019] For example, the electronic device can be a mobile phone, a tablet computer, a wearable device, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), a desktop computer, a smart large screen, a smart TV, a handheld device with wireless communication function, a computing device or other processing devices connected to a wireless modem, a computer, a laptop computer, a customer premise equipment (CPE), and / or other devices for communicating on a wireless system, and a next-generation communication system. For example, a mobile terminal in a 5G network or a mobile terminal in a future evolved Public Land Mobile Network (PLMN), etc.

[0020] To better understand the group teaching strategy generation method based on the intelligent learning cloud platform provided in the embodiments of the present application, the following provides an exemplary introduction to the specific implementation process of the group teaching strategy generation method based on the intelligent learning cloud platform provided in the embodiments of the present application.

[0021] Figure 1 The schematic flowchart of the group teaching strategy generation method based on the intelligent learning cloud platform provided in the embodiments of the present application is shown. The group teaching strategy generation method based on the intelligent learning cloud platform includes: S100, obtaining teaching behavior data, biological data, and teaching progress data of multiple classes through the intelligent learning cloud platform. Among them, the teaching progress data includes the taught knowledge points and the corresponding teaching dates.

[0022] It can be understood that the teaching behavior data can include students' test records, interaction trajectories with the learning platform, and social interaction data. The test records can include students' scores and answering situations in various online tests, which can reflect students' mastery of knowledge points; the interaction trajectories with the learning platform can include students' interactive behaviors such as clicking on course materials, browsing learning resources, and participating in online discussions; the social interaction data can include students' speeches and comments in the discussion area, or their interaction situations in collaborative tasks. The social interaction data can reflect students' communication and interaction with classmates or teachers, and help understand students' teamwork spirit, expression ability, and participation in group cooperation.

[0023] Exemplarily, teaching behavior data can be automatically collected through the intelligent learning cloud platform. The intelligent learning cloud platform can record every learning activity of students and record the data of each student's interaction with the platform through a database.

[0024] It can be understood that biological data can be collected through biosensor devices or other sensing tools, which can provide information about students' physiological responses, emotions, and psychological states, and help understand situations such as stress, attention, and mood swings that students may experience during the learning process. Biological data can include heart rate variability (HRV), galvanic skin response (GSR), facial expression recognition, and microphone speech prosody. Heart rate variability (HRV) can reflect students' mood swings, stress levels, and physical health conditions by measuring the changes in students' heart rates; galvanic skin response (GSR) can evaluate students' emotional responses through the changes in skin conductance, and GSR can monitor mood changes such as anxiety, excitement, or tension; facial expression recognition can evaluate the emotional responses presented by students during the learning process through facial expression analysis technology; microphone speech prosody can evaluate students' emotional states and psychological responses by analyzing the speech features (such as pitch, speech rate, and intonation changes) when students speak.

[0025] Exemplarily, biological data can be collected in real time through specific hardware devices (such as smartwatches, health monitoring devices, facial recognition cameras, etc.) and transmitted to the intelligent learning cloud platform for storage and analysis through wireless transmission methods such as Bluetooth, Zig-Bee, and WiFi. For example, heart rate monitoring can be completed by a worn smart bracelet or watch, galvanic skin response can be recorded through wearable sensors, facial expression recognition can be analyzed in real time through cameras and computer vision technology, and speech prosody data can be captured by a microphone and analyzed for speech signals.

[0026] It can be understood that teaching progress data is the record of the teaching process of a course by teachers, including the taught knowledge points, teaching content, and the teaching dates corresponding to each knowledge point, which can help teaching managers and teachers track the course progress and arrangements and ensure that the teaching content of all classes can be covered on time.

[0027] Exemplarily, teaching progress data can be manually input by teachers or automatically recorded through the intelligent cloud learning platform, the teaching dates of each knowledge point can be marked, and the progress data can be updated. It can be displayed in the form of a course calendar or a course progress schedule, or stored and accessed through reports generated by the system.

[0028] The teaching behavior data, biological data, and teaching progress data stored through the intelligent learning cloud platform can not only comprehensively understand students' learning states and emotional changes, but also help teachers conduct personalized teaching, optimize group teaching strategies, and provide accurate data support for teaching decisions.

[0029] S200 performs spatio-temporal alignment processing on the teaching stages of multiple classes based on the teaching progress data of the multiple classes to obtain a teaching stage sequence. Among them, the teaching stage sequence includes multiple teaching stages, and each teaching stage includes the teaching behavior data and biological data of multiple classes in the same teaching stage.

[0030] It can be understood that the teaching stages can be divided according to the knowledge point teaching dates in the teaching progress data, and the start and end of each stage can be defined by the teaching dates of a series of teaching contents. For example, a teaching stage may include the teaching contents from knowledge point A to knowledge point B.

[0031] The teaching progress data of all classes can be unified to a standard time scale, such as using weeks or days as units, which is conducive to horizontal comparison of the teaching progress between different classes. Teaching stage alignment can solve the problem of data comparability caused by differences in the teaching progress of different classes.

[0032] Exemplarily, for the teaching progress of different classes, due to the progress differences between classes, the teaching start dates of different classes can be adjusted. Assume that the starting points of the first stage of all classes are aligned based on a certain standard date (such as the first stage of all classes starts from the first week), and then each class synchronizes the time of each teaching stage according to its actual progress arrangement.

[0033] Each class may have different teaching contents and sequences. Spatial alignment means aligning the teaching contents of each class into a unified standard sequence according to their teaching progress arrangements. For example, a certain class may teach knowledge point A in the second stage, while another class may teach knowledge point B in the second stage. Therefore, the teaching contents of each class can be uniformly divided into stages based on the teaching progress data, so that each class teaches similar contents within the same teaching stage.

[0034] After the spatio-temporal alignment processing is completed, the teaching behavior data and biological data of each class are grouped and classified according to each teaching stage to form an overall teaching stage sequence. Each teaching stage contains the teaching behavior data and biological data of all classes in this teaching stage. Each teaching stage can form a data matrix or sequence, where each element represents the data of a certain class in this stage, and these data will be used for subsequent analysis, learning progress evaluation, and personalized learning intervention.

[0035] This step can help teachers and administrators understand the learning dynamics of each class in different teaching stages, and at the same time provides strong data support for personalized learning, teaching strategy optimization, and learning situation analysis.

[0036] In a possible implementation, in S200, according to the teaching progress data of multiple classes, the teaching stages of multiple classes are processed for spatio-temporal alignment to obtain a teaching stage sequence, including: S210, compare whether the knowledge points in the teaching progress data of multiple classes are consistent with the knowledge points in the standard teaching framework. If they are inconsistent, according to the equivalent knowledge point table, replace the inconsistent knowledge points in the teaching progress data with the corresponding knowledge points in the standard teaching framework. Among them, the equivalent knowledge point table includes the equivalent relationships between the same or similar knowledge points in different textbooks and the knowledge points in the standard teaching framework. The equivalent knowledge point table is obtained through semantic similarity calculation and manual rule library supplementation. The standard teaching framework includes each teaching stage ID, the knowledge points included in each teaching stage, and the expected teaching duration of each teaching stage.

[0037] It can be understood that each class may use different textbooks (it is possible that different schools are all using the intelligent learning cloud platform), and the expressions of the same knowledge point in textbooks may be different (for example, in textbook A, the straight-line equation may be taught, while in textbook B, the image of a linear function is taught). The corresponding knowledge points in the standard teaching framework for different expressions may be the same or equivalent. The equivalent knowledge point table is to solve the knowledge point differences between different textbooks and correspond the knowledge points in different textbooks to the knowledge points in the standard teaching framework.

[0038] Exemplarily, when aligning the class progress, the textbooks and knowledge points used in the actual teaching of the class can be detected to check whether they are exactly the same as the knowledge points in the standard teaching framework. If the knowledge points in the textbook used by the class are not completely consistent with the knowledge points in the standard teaching framework, then they can be replaced with the corresponding knowledge points in the standard teaching framework according to the equivalent knowledge point table. For example, if a certain knowledge point in the textbook used by class A is the straight-line equation, and the corresponding image of a linear function in the standard teaching framework is found through the equivalent knowledge point table, then when aligning, the straight-line equation learned by class A is replaced with the image of a linear function in the standard teaching framework. Similarly, if the calculation of Ohm's law in the textbook used by class B is not completely consistent with the circuit law in the standard teaching framework, then according to the equivalent knowledge point table, the calculation of Ohm's law in class B will also be replaced with the knowledge points in the standard teaching framework, which is beneficial to the consistency of alignment.

[0039] By using the equivalent knowledge point table, it can be ensured that regardless of what textbooks the classes use, the teaching progress can ultimately be accurately mapped to the corresponding knowledge points in the standard teaching framework, which is beneficial to eliminating textbook differences and unifying the expression of knowledge points when aligning the progress of different classes.

[0040] S220. Calculate the distance matrix between the knowledge points in the teaching progress data of each class and the knowledge points in the standard teaching framework; based on each distance matrix, use the dynamic time warping algorithm to calculate the optimal alignment path between the teaching progress of each class and the standard teaching framework; according to the optimal alignment path between the teaching progress of each class and the standard teaching framework, map the teaching progress of each class to each teaching stage in the standard teaching framework to obtain the teaching stage sequence.

[0041] It can be understood that the dynamic time warping algorithm (DTW) is an algorithm used to compare the similarity of two sequences, which can handle sequences of different lengths and different speeds and find the optimal alignment path. In teaching progress alignment, DTW can warp the teaching progress sequences of different classes to make up for the time differences. Even if the progress between classes is different, DTW can find the alignment path that minimizes the cost and match the knowledge points of each class to the corresponding knowledge points in the standard teaching framework. DTW allows stretching and shrinking of sequences. Even if a class may learn some knowledge points earlier or later, DTW can align these differences. Through insertion or deletion steps, the progress sequences of different classes can coincide at some knowledge points.

[0042] Exemplarily, the distance matrix between the knowledge points in the teaching progress of each class and the knowledge points in the standard teaching framework can be calculated. The distance matrix can be measured based on factors such as timestamp differences and the similarity of knowledge points (such as semantic distance). The distance matrix reflects the matching degree between the class teaching progress and the standard teaching framework.

[0043] The DTW algorithm can be used to find the optimal alignment path between the teaching progress of each class and the standard teaching framework. The optimal alignment path is completed by calculating the minimum cost path for each point in the distance matrix. DTW allows stretching and shrinking in time, that is, the progress of a class can be advanced or postponed to achieve the minimum cost, that is, the most reasonable alignment method. For example, class A may have explained the function definition at a certain moment, while class B has not reached this part yet. DTW can flexibly adjust the progress of class B, stretching or postponing the teaching time of the function definition.

[0044] After finding the optimal alignment path, the teaching progress of each class will be mapped to each teaching stage of the standard teaching framework. The DTW algorithm can generate a teaching stage sequence, showing which stage of the standard teaching framework the actual teaching time of each class corresponds to. The teaching behavior data and biological data of each class can be grouped and classified according to each teaching stage based on the timestamps in the teaching behavior data and biological data of each class.

[0045] Through the DTW algorithm, elastic alignment can be achieved among multiple classes, progress differences can be handled, and the teaching progress of multiple classes can be effectively aligned to the standard teaching framework, which is conducive to reasonably handling the progress differences among different classes and making the data between classes comparable.

[0046] In a possible implementation, the method for generating a group teaching strategy based on the intelligent learning cloud platform further includes: S201, extracting knowledge points from the teaching content based on text analysis technology. The teaching content is pre-stored in the intelligent learning cloud platform.

[0047] Exemplarily, the teaching content exists in a structured or semi-structured text form. The text data can be preprocessed for subsequent analysis. The long teaching content can be split into paragraphs and sentences so that subsequent analysis can be more clearly targeted at specific content; a Chinese word segmentation tool (such as HanLP, LTP, Jieba, etc.) can be used to segment the text into smaller units (such as words or phrases); stop words in the text can be removed to reduce noise and improve analysis efficiency. Stop words refer to words that frequently appear in text analysis but have no actual meaning (such as "de", "shi", etc.); each word in the text can be classified by part of speech (such as noun, verb, adjective, etc.). When identifying knowledge points, noun phrases (such as "physical laws", "triangle") are given priority as potential knowledge points.

[0048] Key entities in the text can be identified through named entity recognition (NER) technology. For example, core concepts (such as Newton's laws, particle motion) and methods (such as derivative rules) in disciplines such as physics, mathematics, and chemistry. Topic modeling technology (such as Latent Dirichlet Allocation, LDA) can be used to automatically discover potential topics or core concepts from the teaching content, thereby helping to identify knowledge points. Important keywords in the teaching content can be extracted through algorithms such as TF-IDF and TextRank, and these keywords may be core knowledge points.

[0049] S202, analyzing the context in the teaching content using a relationship extraction algorithm, identifying the dependency relationships between knowledge points, and constructing a knowledge point dependency graph.

[0050] Exemplarily, dependency parsing is an important tool for relationship extraction, which can identify the dependency relationships between words. Each sentence forms a syntactic tree, where each word is a node and the edge represents the dependency relationship between words. For example, in the statement that acceleration is proportional to force, there is a dependency relationship between acceleration and force, indicating that acceleration depends on force. Dependency analysis can identify the syntactic associations between these knowledge points and construct a higher-level understanding.

[0051] In addition to syntactic relations, the relationships between more complex knowledge points can also be identified through semantic analysis. Semantic relation extraction involves understanding the deep meaning of the text, not just the grammatical relations. For example, causal relations (such as an increase in heat causing an increase in the temperature of an object), prerequisite relations (such as mastering basic mathematics to understand advanced mathematics), and progressive relations (such as deriving Newton's second law from Newton's first law).

[0052] Once the dependencies between knowledge points are identified, the dependency graph of knowledge points can be constructed using these dependencies. The dependency graph of knowledge points is a directed graph, where nodes represent knowledge points and edges represent the dependencies between knowledge points, i.e., prerequisite knowledge point -> current knowledge point.

[0053] By analyzing the context in the teaching content through a relation extraction algorithm, the dependencies between various knowledge points in the teaching content can be identified. These dependencies include not only syntactic connections but also semantic-level causal, prerequisite, progressive, etc. relations. The dependency graph of knowledge points can organize knowledge points into an ordered structure, helping to implement tasks such as teaching design, learning recommendation, and curriculum optimization.

[0054] S203. According to the dependency graph of knowledge points, divide the knowledge points into each teaching stage and label an ID for each teaching stage.

[0055] Exemplarily, a topological sort can be performed on the dependency graph of knowledge points. The role of topological sorting is to be able to determine the teaching order of knowledge points, which is beneficial for all prerequisite knowledge points to have been taught before teaching a certain knowledge point. Calculate the in-degree of each node (i.e., the number of edges pointing to the node), add all nodes with an in-degree of 0 to the queue (these nodes have no prerequisite dependencies and can be taught first), take out nodes from the queue one by one, and decrement the in-degree of the nodes they depend on by 1. If the in-degree of a dependent node becomes 0, add that node to the queue. Continue the calculation according to the above steps until all nodes are processed. Through topological sorting, an ordered list that satisfies the dependencies between knowledge points can be obtained.

[0056] When the topological sorting is completed, simpler knowledge points and more complex knowledge points can be assigned to different stages according to the difficulty of the knowledge points or the teaching objectives. For example, basic concepts, laws, formulas, etc. can be assigned to the early stages, while complex derivation processes, application methods, etc. can be assigned to the later stages. According to the order of topological sorting, the knowledge points can be divided into multiple teaching stages according to the dependency relationship. Knowledge points with more dependencies can be arranged in subsequent stages, while basic knowledge points should be arranged in the early stages.

[0057] After the teaching stages are divided, a unique ID can be assigned to each teaching stage. The ID can be assigned using a simple numerical numbering. For example, teaching stage S1 includes knowledge points K1 and K2, teaching stage S2 includes knowledge points K3, K4, and K5, and teaching stage S3 includes knowledge points K6 and K7.

[0058] The division of teaching stages is conducive to students gradually mastering knowledge, and the dependency relationships between knowledge points ensure that the teaching content of each stage is built on the basis of the previous stage.

[0059] S204, Based on the number of knowledge points in each teaching stage, allocate the expected teaching duration for each teaching stage.

[0060] Exemplarily, the average teaching time required for each knowledge point can be calculated according to historical teaching data. According to the average teaching time required for each knowledge point, calculate the teaching time required for all knowledge points in each teaching stage, and determine the teaching time required for all knowledge points in each teaching stage as the corresponding expected teaching duration.

[0061] S205, Arrange each teaching stage ID, the knowledge points included in each teaching stage, and the expected teaching duration of each teaching stage according to the dependency relationships between knowledge points to obtain a standard teaching framework.

[0062] Exemplarily, according to the topological sorting result of the knowledge point dependency relationship graph, output an ordered list of teaching stages (standard teaching framework), including each teaching stage ID, the knowledge points included in each teaching stage, and the expected teaching duration of each teaching stage.

[0063] Through the above steps, support can be provided for aligning the teaching progress of subsequent classes.

[0064] In a possible implementation, the method for generating a group teaching strategy based on an intelligent learning cloud platform further includes: S2001, During the spatio-temporal alignment process, use the dynamic time warping algorithm to detect whether there are progress jump features. Among them, the progress jump feature means that the knowledge points corresponding to the teaching progress of a certain class in the standard teaching framework are not covered by the teaching progress data of that class.

[0065] Exemplarily, when using the DTW algorithm to align the teaching progress of a class, if a class skips a knowledge point (such as directly entering the equation solution method without learning the nature of equations), the DTW algorithm can find that the knowledge points included in the standard teaching framework are not covered by the actual progress of that class.

[0066] For example, in the alignment path, the knowledge points corresponding to the class progress in the standard teaching framework (such as the properties of equations) do not match any teaching time points of the class. The actual progress of the class directly skips the prerequisite knowledge points (such as the properties of equations) and starts learning the content of the subsequent stage (such as the solution methods of equations).

[0067] To handle the situation of progress jumps, virtual teaching stages can be inserted or the missing knowledge points can be marked.

[0068] S2002, if there is a progress jump feature, insert virtual knowledge points at the positions of the knowledge points not covered in the teaching stage sequence to fill in the uncovered knowledge points. Or, if there is a progress jump feature, mark the knowledge points not covered in the teaching stage sequence and indicate that the uncovered knowledge points are missing.

[0069] Exemplarily, when the DTW algorithm detects that there are progress jumps in some classes, the skipped knowledge points can be marked, and a virtual knowledge point can be inserted at the position of the skipped knowledge points in the teaching stage sequence, marking this knowledge point as to be supplemented. The virtual knowledge point is used to fill in the missing content and does not represent actual teaching content, but can remind that this part of knowledge needs to be supplemented in subsequent teaching arrangements.

[0070] For example, if class A directly enters the solution methods of equations, skipping the properties of equations, then class A in the teaching stage sequence can be as shown in Table 1: Table 1 Another method to handle progress jumps is to mark the missing knowledge points, mark the skipped knowledge points in the teaching stage sequence, and indicate the missing. Teachers can supplement the missing knowledge points according to the marks in subsequent teaching plans, which is conducive to students mastering the complete course content.

[0071] For example, if class A skips the properties of equations and directly starts the solution methods of equations, then class A in the teaching stage sequence can be as shown in Table 2: Table 2 Through the missing knowledge points in the spatio-temporal alignment results, reminder information can be generated to prompt teachers to supplement the missing knowledge points in the subsequent teaching process. Teachers can view the reminder information on the intelligent learning cloud platform. This not only helps teachers discover omissions in a timely manner but also contributes to the integrity of the teaching stage sequence.

[0072] The alignment of teaching stages uses three core technologies: constructing a standard teaching framework, mapping flexible progress, and semantic matching of knowledge points. It converts scattered and heterogeneous teaching data into normalized inputs that can be compared in terms of time and space, providing a reliable basis for subsequent group anomaly detection and strategy generation. This process not only preserves the original teaching characteristics but also eliminates the analysis bias caused by progress noise.

[0073] S300, for each teaching stage in the teaching stage sequence, perform group anomaly detection on multiple classes to obtain the anomaly detection results. Among them, the anomaly detection results include the teaching stages with group anomalies.

[0074] It can be understood that group anomaly detection refers to identifying the abnormal stages based on the teaching behavior data and biological data of all classes in each teaching stage. By comparing the teaching behavior data and biological data of all classes, it is possible to find out which teaching stages have anomalies at the group level, thereby providing a basis for adjusting teaching strategies.

[0075] Exemplarily, there may be missing data in the teaching stage sequence. For example, the biological data or teaching behavior data of some classes may not be recorded in some stages. It can be processed by interpolation, filling, or deleting missing values; if some data is significantly abnormal (such as extreme values of biological data, or illogical high-frequency clicks in teaching behavior data), it can be removed or corrected.

[0076] The Z-score standardization or Min-Max normalization method can be used to standardize or normalize the teaching behavior data and biological data of different classes, so that data with different dimensions can be compared on the same scale.

[0077] The Z-score of the teaching behavior data and biological data of each class in each teaching stage can be calculated, and based on the Z-score value, it can be judged whether it is abnormal. If the Z-score of a certain class is greater than a certain threshold (which can be ±2 or ±3), then this class may have an anomaly in this stage.

[0078] Use clustering algorithms (such as K-means, DBSCAN, etc.) to divide classes into different groups and identify groups that are significantly different from other groups. Through cluster analysis, abnormal behaviors or abnormal patterns within the group can be detected.

[0079] The clustering algorithm (such as K-means, DBSCAN, etc.) can be used to cluster the teaching behavior data and biological data of all classes. The clustering algorithm can capture the common patterns within the group and identify the outliers that are significantly different from most classes. For example, in a certain teaching stage, if the behavior patterns of some classes are significantly different from those of other classes, then these classes may belong to an abnormal group.

[0080] Through the above Z-Score method or clustering algorithm, teaching stages that significantly deviate from the normal trend in a group can be identified, abnormal teaching stages can be marked, and whether there are certain common problems can be analyzed through the biological data and teaching behavior data of the abnormal teaching stages. For example, it may be that the effect of a certain teaching method is not good, a certain knowledge point is too complex leading to student anxiety, or the teaching rhythm is inappropriate leading to a decline in student attention.

[0081] The purpose of group anomaly detection is to analyze the teaching behavior data and biological data of multiple classes, discover abnormal manifestations in the teaching stage, and thus provide a basis for adjusting teaching strategies. By reasonably selecting detection methods, combining data preprocessing and anomaly assessment, potential problems in the teaching process can be effectively identified.

[0082] In one possible implementation, refer to FIG. 2, S300. Perform group anomaly detection on multiple classes for each teaching stage based on the teaching stage sequence to obtain anomaly detection results, including: S310, Extract the teaching behavior data and biological data of multiple classes in each teaching stage from the teaching stage sequence.

[0083] Exemplarily, group anomaly detection targets each teaching stage and is used to detect whether there is a group anomaly situation in each teaching stage (such as a very high error rate of multiple classes on the same problem). Each teaching stage in the teaching stage sequence already contains the teaching behavior data and biological data of all classes, and the teaching behavior data and biological data in each teaching stage can be analyzed to detect whether there is a group anomaly in this teaching stage.

[0084] Perform the following processing for each teaching stage: S320, Extract common error patterns from the teaching behavior data of multiple classes; construct a cognitive network diagram based on the common error patterns, and based on the cognitive network diagram, identify group knowledge breakpoints; integrate the common error patterns and group knowledge breakpoints to obtain group cognitive characteristics. Among them, the cognitive network diagram is used to represent the mastery situation and correlation of different knowledge points by students during the learning process.

[0085] Exemplarily, through cluster analysis, the error patterns that students often encounter during the learning process can be identified. By clustering the error data in the teaching behavior data by knowledge points, it can be identified which knowledge points have common errors in multiple classes. For example, the concepts of mass and weight are confused. By analyzing students' homework, quizzes, and question-and-answer records, these error patterns can be identified, and further inferences can be made about which knowledge points may be the root causes of these errors.

[0086] Each knowledge point is extracted from the teaching content and serves as a node in the cognitive network graph. A node contains the name, definition, related concepts, and mastery level of a knowledge point. The relationships between knowledge points can be identified by analyzing the commonalities in error patterns. For example, if students often make mistakes when learning the laws of mechanics, and these mistakes affect their subsequent learning of energy conversion, then there is a dependency relationship between these two knowledge points. Based on the dependency relationship between knowledge points and the frequency of common mistakes, the weight of the edge can be calculated. The greater the weight, the stronger the dependency relationship between knowledge points, and the learning of one knowledge point by students may affect the learning of other knowledge points. By identifying the dependency relationship and calculating the common error patterns between knowledge points, a weighted directed graph (cognitive network graph) is established. Each node in the graph represents a knowledge point, and the weight of each edge represents the similarity of students' mastery of related knowledge points.

[0087] Whenever new test or learning data is fed back, the cognitive network graph can be automatically updated. The mastery of knowledge points by students at different stages can affect the weights and connection relationships of knowledge points in the cognitive network graph.

[0088] In the cognitive network graph, knowledge points with a low mastery level can be marked as potential breakpoints. These potential breakpoints are knowledge points that students generally do not master well during the learning process and may be the core areas of cognitive breaks. The impact of low-mastery knowledge points on other related knowledge points can be calculated through propagation algorithms. For example, if students generally make mistakes in a certain knowledge point (such as the basic formulas of mechanics), and this mistake affects subsequent knowledge points (such as the understanding of the law of conservation of energy), then this knowledge point and the subsequent knowledge points will be marked as group knowledge breakpoints. The most influential breakpoints in the group can be calculated, that is, those knowledge points that have a wide impact on the learning path. By analyzing the connectivity of the cognitive network graph and the correlation between knowledge points, it can be determined which lack of mastery of knowledge points leads to subsequent cognitive faults.

[0089] The identified common error patterns can be combined with group knowledge breakpoints to form group cognitive characteristics. Through group cognitive characteristics, it is possible to know the common mistakes of students in certain knowledge points, common understanding obstacles, and which knowledge points are the learning bottlenecks of the entire group, providing a basis for optimizing group teaching strategies.

[0090] S330, Based on the teaching behavior data of multiple classes, use Fourier transform to perform frequency analysis on students' learning behaviors, identify high-frequency and low-frequency components, and obtain the behavior frequency spectrum; calculate the resonance matching degree between the behavior frequency spectrum and the ideal learning path frequency spectrum; identify non-explicit behavior patterns based on the resonance matching degree to obtain group behavior characteristics.

[0091] It can be understood that step S330 is the implementation process of the cross-modal behavior resonance analysis method. By combining physical field theory, the teaching behavior data of students is regarded as vibration signals in an energy field, and then the matching degree with the ideal learning path is analyzed. In traditional behavior analysis, the operations of students are regarded as simple activities or tasks. However, cross-modal behavior resonance analysis borrows the concepts of energy field and vibration frequency in physical field theory, regards the behavior of students as a dynamic system, and uses tools such as Fourier transform to transform the behavior pattern into a frequency spectrum, enabling the identification of behavior resonance phenomena in the learning process of students and optimizing the group teaching strategy based on this.

[0092] The behavior frequency spectrum is similar to the frequency spectrum of a physical vibration system. The operation rhythm of students can be analyzed through Fourier transform to extract the main frequency components in the operation.

[0093] The resonance matching degree is to compare the frequency spectra between the current behavior pattern of students and the ideal learning path, and calculate the harmonic coincidence degree, that is, whether the behavior pattern resonantly matches the ideal path. The level of the resonance matching degree reflects whether the learning state of students is consistent with the ideal learning state.

[0094] Exemplarily, time-domain data (the change of students' behavior over time) can be converted into frequency-domain data through Fourier transform, that is, by analyzing the frequency components of the signal to capture periodic behavior patterns. The behavior frequency spectrum obtained after Fourier transform represents different frequency components in students' learning behavior, and the behavior frequency spectrum reveals the periodic law of students' learning behavior. High-frequency components represent short-term rapid change behaviors in the learning process of students (such as temporary anxiety, sudden attention fluctuations, etc.), and high frequencies may be related to short-term behavior responses, task completion, etc.; low-frequency components represent long-term stable behaviors in the learning process of students (such as regular learning, long-term attention persistence, etc.), and low-frequency components reflect the long-term trend or stable behavior pattern of students' learning.

[0095] The ideal learning path can be defined by the best learning rhythm of students, the order of knowledge point mastery, and the depth and coherence of learning. For example, in the ideal learning path, students switch tasks regularly and progress step by step, maintaining focus and in-depth thinking. Using Fourier transform to perform frequency analysis on the ideal learning path, its frequency spectrum is relatively stable and regular.

[0096] The resonance matching degree between the student behavior frequency spectrum and the ideal learning path frequency spectrum can be calculated through cross-correlation analysis or similarity measurement to measure the similarity between the behavior frequency spectrum and the ideal learning path frequency spectrum. Cross-correlation analysis can quantify the correlation between them by calculating the cross-correlation function of the student behavior frequency spectrum and the ideal learning path frequency spectrum; similarity measurement can use methods such as cosine similarity and Euclidean distance to evaluate the similarity between the two. If the learning behavior frequency of the student matches the ideal learning path frequency relatively high, it indicates that the student is in an ideal learning state; if the matching degree is low, it indicates that the student may be in an inefficient learning state and needs intervention.

[0097] Non-explicit behavior patterns refer to problems that are not obviously manifested in the learning process of students but can be revealed from the resonance matching degree of the frequency spectrum. These patterns may be some subtle and implicit behaviors of students in learning, such as frequent attention fluctuations, short-term anxiety, unstable learning habits, etc. The behavior frequency spectrum and resonance matching degree can be analyzed to identify abnormal points in the student behavior frequency spectrum and infer these non-explicit behavior patterns.

[0098] Based on the teaching behavior data of multiple classes, the behavior frequency spectra of all students can be analyzed, and the resonance matching degree of each student can be calculated. Through cluster analysis or statistical methods, the behavior frequency spectra of students are divided into several groups to identify the overall learning patterns of the groups. For the students within the group, the commonly existing behavior characteristics are identified. For example, most students show a stable learning pattern (mainly low-frequency spectra), and some classes of students frequently have high-frequency behavior fluctuations, which may indicate cognitive or emotional problems.

[0099] Through this step, the states of students' attention dispersion, deep learning, etc. can be identified, and teaching strategies can be dynamically adjusted according to the behavior patterns of students. Through the cross-modal behavior resonance analysis method, the learning states of students can be understood more accurately, and teaching support can be provided targeted, thereby improving the teaching effect.

[0100] S340, calculate the stress index according to the biological data of multiple classes to obtain the group emotional characteristics.

[0101] Exemplarily, the stress index of each student can be calculated by using the weighted summation method for four biological signals (heart rate variability (HRV), galvanic skin response (GSR), facial expression recognition, microphone speech prosody) of each student.

[0102] The mean or weighted average of the stress indices of all students within each class can be calculated to obtain the overall stress index of each class, that is , where represents the class stress index of the jth class, $S_{i}$ represents the stress index of the $i$-th student in the class, and $N$ represents the number of students in the class.

[0103] By calculating the stress indices of all classes, the overall emotional fluctuations (group emotional characteristics) of the group can be obtained. The mean value of the stress indices of all classes can be calculated to reflect the group stress level, that is where $P$ represents the group stress level (average value), and $M$ represents the number of classes. The standard deviation of the stress indices of all classes can be calculated to reflect the group emotional fluctuations, that is where $E$ represents the group emotional fluctuations (standard deviation). The mean value and standard deviation of the stress indices of all classes are determined as the group emotional characteristics.

[0104] S350. Perform feature fusion on the group cognitive characteristics, group behavior characteristics, and group emotional characteristics to obtain the fused characteristics.

[0105] Exemplarily, corresponding weights can be assigned to the group cognitive characteristics, group behavior characteristics, and group emotional characteristics, and weighted averaging is performed on the group cognitive characteristics, group behavior characteristics, and group emotional characteristics according to the weights corresponding to the three characteristics to obtain the fused characteristics.

[0106] If each feature dimension is high and the redundancy is strong, principal component analysis (PCA) can be used for dimensionality reduction. Multiple features can be synthesized into one or several principal components, which can reduce the dimension of the data and retain most of the information.

[0107] A machine learning model can be trained to automatically learn the optimal feature fusion method, and the model can automatically determine the fusion method and importance of different features according to the relationships in the data.

[0108] The fused characteristics can be used as the characteristics of the overall learning state of the group for subsequent anomaly detection and teaching strategy generation.

[0109] S360. Compare the fused characteristics with the dynamic anomaly threshold. If the fused characteristics are greater than the dynamic anomaly threshold, it is determined that there is a group anomaly in this teaching stage, and the anomaly detection result is obtained.

[0110] Exemplarily, compare the fused characteristics with the dynamic anomaly threshold. If the value of the fused characteristics is greater than the dynamic anomaly threshold, it is determined that there is a group anomaly in this teaching stage. Or statistical methods or machine learning models (such as anomaly detection algorithms) can be used to identify group behavior patterns that exceed the normal range.

[0111] If the value of the fused characteristics exceeds the threshold, this teaching stage can be marked as an abnormal stage, and corresponding alarms or reports can be generated. Abnormal teaching stages may include emotional fluctuations of the student group, a sharp increase in the cognitive error rate, or behavior deviation from the normal learning path, etc.

[0112] By analyzing historical data, the normal range of the group can be calculated, and abnormal thresholds can be calculated through statistical methods (such as standard deviation, quantiles, etc.). As new teaching behavior data and biological data are input, the dynamic abnormal threshold can be updated according to real-time data, which helps the threshold to be self-adaptive.

[0113] Through this step, it is possible to identify and feedback abnormal situations of the group in the learning process in real time, so as to support personalized teaching and timely intervention.

[0114] In a possible implementation manner, the method for generating group teaching strategies based on the intelligent learning cloud platform further includes: MAML (Meta-Learning Model) is a model-agnostic meta-learning algorithm, aiming to enable the learned model parameters to quickly adapt when encountering new tasks through a small amount of task training. The key of MAML is to train on multiple tasks so that the model can learn an initial parameter, which can quickly adapt to different tasks through a small number of gradient updates.

[0115] S301, Obtain the teaching behavior data of all classes in multiple historical teaching stages, and determine the teaching behavior data of all classes in each historical teaching stage as a meta-task.

[0116] Exemplarily, the teaching behavior data of all classes in a certain grade in previous years can be obtained from the intelligent learning cloud platform. In the MAML framework, each task represents a learning task or learning scenario, and the behavior data of all classes within a teaching stage are regarded as the same meta-task. The meta-task contains the teaching behavior data of all classes in this stage. Each teaching stage corresponds to a meta-task, but the data of different teaching stages will be trained and evaluated as different meta-tasks. The training of all teaching stage data can jointly optimize the initial parameters of the model, enabling the model to quickly adapt to all classes in each teaching stage and extract the common behavior characteristics of all classes.

[0117] S302, For each meta-task, divide the teaching behavior data of all classes into a training set and a test set. Train the meta-learning model based on the training set, and calculate the loss value of the meta-learning model on the training set. According to the loss value, update the model parameters of the meta-learning model through the gradient descent algorithm. Calculate the test loss based on the test set, and calculate the gradient of the test loss with respect to the model parameters based on the test loss.

[0118] Exemplarily, a neural network model can be defined, and this model can learn the common behavior characteristics of all classes in a certain teaching stage. The initial parameters of the model It will be optimized so that the model can quickly adapt to new class data with a small number of updates. During each training, MAML randomly selects a meta-task (i.e., a teaching phase) from the data of all teaching phases, and then uses all the class data within that meta-task for training and testing.

[0119] For each meta-task (each teaching phase), a part of the data within that teaching phase can be used as the training set, and the remaining data as the test set. The meta-learning model is trained using the training set within the task (i.e., a certain teaching phase). Forward propagation is performed on this training set, and the loss value of the model on the training set is calculated. Then, according to the loss value, the model parameters are updated using the gradient descent algorithm. This is a local update, and the goal is to make the model adapt to the training data of the current task. After the update is completed, the updated model is evaluated using the test set within the task, the test loss of this meta-task on the test set is calculated, and the gradient of the test loss with respect to the initial model parameters is calculated. The test loss reflects the adaptation ability of the meta-learning model after training and can check whether the model can adapt well to new data.

[0120] S303, calculate the average value of the gradients of all meta-tasks to obtain the meta-gradient. Update the model parameters of the meta-learning model according to the meta-gradient to obtain the meta-learning model after training is completed.

[0121] Exemplarily, according to the gradients of the test losses of all tasks with respect to the initial model parameters calculate the average value of these gradients to obtain the meta-gradient. The meta-gradient reflects the generalization ability of the meta-learning model on multiple meta-tasks and helps the model learn an initial parameter that can adapt to all teaching phases. Update the initial parameters of the meta-learning model through the meta-gradient so that this initial parameter can quickly adapt when facing new tasks (such as a new teaching phase).

[0122] S304, input the teaching behavior data of multiple classes in each teaching phase into the meta-learning model after training is completed, so that the meta-learning model after training can extract the behavior gradients and balance coefficients of each teaching phase, and determine the behavior gradients and balance coefficients of each teaching phase as the group behavior characteristics of each teaching phase.

[0123] Exemplarily, during the training process, through meta-task inner gradient update, the adjustment direction and amplitude of the initial model parameters for each meta-task can be obtained. Through meta-update, the model can capture the behavioral gradients at each teaching stage, that is, the strategy adjustment directions of different classes during the learning process. By analyzing the task switching frequency, time allocation, etc. in the teaching behavior data, an exploration-exploitation balance coefficient can be extracted, which characterizes the proportion of knowledge deepening and new field exploration of students during the learning process. For example, if a student repeatedly practices certain knowledge points, it indicates that the student tends to exploit; if a student often switches tasks, it indicates that the student tends to explore.

[0124] After the meta-learning model training is completed, the teaching behavior data of each teaching stage in the teaching stage sequence can be input into the meta-learning model, and the meta-learning model can extract the behavioral gradients and exploration-exploitation balance coefficients of each teaching stage.

[0125] Using the MAML framework for training, shared group behavior characteristics can be extracted from the teaching behavior data of multiple classes, and the common rules of the group's behavior patterns during the learning process can be identified. The training of the MAML model realizes the goal of cross-task learning by optimizing the shared parameters, so that the final group behavior characteristics can reflect the learning characteristics of all classes.

[0126] In a possible implementation manner, the group teaching strategy generation method based on the intelligent learning cloud platform further includes: S3001, calculating the mean and standard deviation of the historical group data, and determining the sum of twice the mean and standard deviation of the historical group data as the initial anomaly threshold. Wherein, the historical group data includes the fusion features of each teaching stage in the teaching stage sequences of multiple historical cycles.

[0127] Exemplarily, the mean and standard deviation of the fusion features of the teaching stage sequences in multiple historical cycles can be calculated, and the initial anomaly threshold can be determined according to the historical mean and historical standard deviation, that is , where represents the initial anomaly threshold, represents the historical mean, represents the historical standard deviation.

[0128] S3002, calculating the standard deviation of the teaching stage sequence based on the fusion features of all teaching stages of the teaching stage sequence.

[0129] Exemplarily, calculating the standard deviation of the teaching stage sequence is to understand the behavioral and emotional fluctuations of the current group during the teaching stage. If the standard deviation is large, it indicates that the group's behavior fluctuates greatly and there may be large abnormal fluctuations; if the standard deviation is small, it indicates that the group's behavior is relatively stable.

[0130] Similarly, according to the integration characteristics of each teaching stage in the teaching stage sequence, calculate the mean of the teaching stage sequence, and calculate the standard deviation of the teaching stage sequence based on the mean.

[0131] S3003. Compare the standard deviation of the historical group data with the standard deviation of the teaching stage sequence to obtain a comparison result, and adjust the multiple of the standard deviation in the initial anomaly threshold according to the comparison result to obtain a dynamic anomaly threshold.

[0132] Exemplarily, the standard deviation of the historical group data can be compared with the standard deviation of the teaching stage sequence. If the standard deviation of the current teaching stage sequence is significantly greater than the standard deviation of the historical group data (i.e., the ratio is large), it indicates that there may be abnormal behaviors or emotional fluctuations in the current teaching stage sequence, and the state of group learning may be in an unstable state; if the standard deviation of the teaching stage sequence is close to or less than the standard deviation of the historical group data, it means that the group state in the current teaching stage sequence is consistent with the historical data and there are no obvious anomalies. The multiple of the standard deviation in the initial anomaly threshold can be dynamically adjusted according to the standard deviation comparison result, that is , where k is a multiple dynamically adjusted according to the comparison result. If the standard deviation of the current teaching stage sequence is significantly greater than the standard deviation of the historical group data, k can be increased to reflect the abnormal fluctuations of the group; if the standard deviation of the current teaching stage sequence is small, k can be decreased, meaning that the group behavior is relatively stable.

[0133] The above steps can dynamically adjust the threshold based on the historical data and the standard deviation of the current stage. It can flexibly adapt to changes in group behavior, cognition, and emotional characteristics, and at the same time enhance the ability to detect anomalies.

[0134] S400. Generate a group teaching strategy based on the teaching behavior data and biological data of the teaching stage in the anomaly detection result.

[0135] Exemplarily, based on the teaching behavior data and biological data in the teaching stage with group anomalies, a machine learning model (such as a decision tree or a support vector machine (SVM)) can be used for causal analysis to determine which factors (such as the teaching of specific knowledge points, the frequency of certain social interactions, etc.) have a significant impact on group anomalies.

[0136] The goal of generating the teaching strategy can be set. For example, if students show a high biological response (such as an increase in skin conductance response) in a certain teaching stage, the goal is to reduce the stress of students and improve concentration; if the interaction frequency of students is low, the goal is to increase classroom interaction and participation; if the grades of students are generally low, the goal is to optimize the teaching content and improve the learning effect of students.

[0137] Group teaching strategies can be automatically generated based on the analysis results and goals through pre-set rules. For example, if biological data (such as galvanic skin response) shows that students have excessive anxiety, strategies such as extending break times, introducing meditation exercises, and reducing teaching difficulty can be generated; if the interaction frequency of students is low, strategies such as increasing the frequency of group discussions, setting more interactive questions, and encouraging online discussions can be adopted; if the academic performance of students in a certain knowledge point is generally low, strategies such as providing more examples, reducing the complexity of knowledge points, and splitting knowledge points into smaller modules can be provided.

[0138] The most suitable group teaching strategies can be dynamically generated based on optimization algorithms such as reinforcement learning or genetic algorithms. Reinforcement learning can simulate different teaching strategies (such as increasing interaction or slowing down the teaching pace) and optimize the strategies according to feedback (such as student performance or emotional data) to finally generate an optimal teaching strategy; genetic algorithms can continuously adjust the parameters of teaching strategies, such as teaching rhythm, interaction frequency, etc., through simulating crossover and mutation until the best strategy combination is found.

[0139] After generating the group teaching strategies, a simulator can be used to evaluate the impact of the strategies on students' behaviors and biological data. If the strategies are effective, the simulation model will show that students' emotions and academic performance have improved. A detailed report can be generated to show the background, goals, proposed measures, and expected effects of the strategies, and the group teaching strategies can be fed back to teachers or educational administrators.

[0140] Through this step, specific group teaching strategies can be automatically generated, which is conducive to the personalization and dynamic optimization of group teaching strategies without relying on human intervention, helping teachers optimize teaching methods and improve students' learning experiences.

[0141] In a possible implementation, please refer to Figure 3 , S400, based on the teaching behavior data and biological data in the teaching stage of the anomaly detection results, generate group teaching strategies, including: S410, select the associated strategies that match the group cognitive characteristics from the strategy knowledge graph; perform associated extension on the common error patterns in the group cognitive characteristics to obtain extended characteristics, and retrieve the extended strategies that match the extended characteristics from the strategy knowledge graph; integrate the associated strategies and extended strategies to obtain the initial teaching strategy.

[0142] Exemplarily, the strategy knowledge graph may include knowledge points, strategies related to teaching strategies, and the relationships between them. For example, each strategy will have some basic attributes (such as applicable knowledge points, implementation methods, teaching goals, etc.), and these strategies can be associated according to different teaching situations and group characteristics.

[0143] Feature matching algorithms (such as keyword-based matching, semantic similarity calculation, etc.) can be used to retrieve strategies from the policy knowledge graph that are related to the common error patterns and knowledge breakpoints in the group cognitive characteristics. When matching strategies, the common error patterns in the group cognitive characteristics can be compared with the defined error types or learning difficulties in the policy knowledge graph for matching. For example, if the common error pattern of the group is formula application error, strategies containing formula application or skill practice can be extracted from the strategy graph. Select the strategies that match the group cognitive characteristics to form a preliminary set of associated strategies.

[0144] The common error patterns in the group cognitive characteristics can be used to identify more knowledge points or potential learning difficulties related to this error pattern through an extension model (such as rule-based reasoning or machine learning model). For example, if there is a common function plotting error in the group, the extended features may involve knowledge points related to image analysis. Generate more learning obstacle features through extension, and these features help to uncover other potential problems that students may encounter in similar situations.

[0145] Based on the extended features, use the feature matching algorithm to retrieve relevant extended strategies from the policy knowledge graph. The extended strategies can involve more extensive or complex teaching methods. For example, to help students better understand through the visual association of images and functions, more practice questions can be designed to promote students' mastery of image analysis skills.

[0146] During the process of merging the associated strategies and extended strategies, deduplication can be performed to facilitate the inclusion of non-duplicate strategies into the initial teaching strategies. Different strategies can be assigned priorities according to the applicability, effectiveness of the strategies or the cognitive difficulties of the group. Finally, an initial teaching strategy will be generated, which contains teaching methods for the current cognitive difficulties, error patterns and extended features of the group.

[0147] S420, based on the group behavior characteristics, adjust the strategy form of the initial teaching strategy, and based on the group emotion characteristics, adjust the strategy attributes of the initial teaching strategy to generate the group teaching strategy.

[0148] Exemplarily, according to the group behavior characteristics, the initial teaching strategy can be formally adjusted by using a dynamic optimization algorithm (such as reinforcement learning) or rule-based adjustment. For example, if the group tends to have low participation, the strategy may need to increase interactivity, and more interactive sessions can be designed and collaborative tasks among students can be increased; if the group behavior shows a low learning progress, the rhythm of the strategy can be adjusted, such as by adding review sessions, extending the task completion time, etc. to adapt to the learning speed of students.

[0149] The subordinate strategies of the initial teaching strategy can be adjusted according to the group's emotional characteristics through a rule engine or machine learning algorithm. For example, if the group's emotions are more anxious or stressed, teaching strategies related to emotional support can be increased, such as relaxation activities and emotional regulation exercises; if the group's emotions fluctuate greatly, the teaching strategy may need to adjust the challenge and rhythm of the tasks to reduce anxiety and improve emotional stability, such as designing tasks with gradually increasing difficulty to avoid negative emotions caused by excessive challenges.

[0150] Integrate the adjustment results of the group's cognitive characteristics, group behavior characteristics, and group emotional characteristics to generate the final group teaching strategy. The group teaching strategy can include teaching content and methods adjusted according to the group's cognitive characteristics, teaching forms and interaction patterns adjusted according to the group's behavior characteristics, and emotional support strategies and learning rhythms adjusted according to the group's emotional characteristics. Output the final group teaching strategy, which can be a specific operation guide for teachers to execute or a strategy recommendation in the intelligent learning cloud platform for further teaching adjustment.

[0151] In a possible implementation manner, the method for generating a group teaching strategy based on an intelligent learning cloud platform further includes: S401, Obtain core relationship pairs and multimodal data sources. Among them, the core relationship pairs include the relationships between misunderstanding patterns, knowledge points, teaching strategies, and teaching resources, and the multimodal data sources include teaching texts and teaching videos.

[0152] Exemplarily, the construction process of the strategy knowledge graph starts from the generation of seed knowledge. The seed knowledge is the core relationship pairs generated through education expert annotation, template mining, or rules, including misunderstanding pattern - knowledge point - teaching strategy - teaching resource. These core relationship pairs can be transformed into structured information and stored as texts, tables, or database records. For example, the core relationship pairs annotated by experts include: misunderstanding pattern: inertial confusion → knowledge point: Newton's first law → teaching strategy: explain the concept of inertia → teaching resource: video of Newton's three laws; misunderstanding pattern: wrong current direction → knowledge point: current direction → teaching strategy: introduce the flow direction of current → teaching resource: exercises on current direction.

[0153] The multimodal data sources can include teaching texts (textbooks, lecture notes, lesson plans, etc.) for extracting knowledge points at the language level; teaching videos (classroom recordings, teaching demonstrations, etc.) for extracting knowledge points at the visual / semantic level; teaching AR (augmented reality teaching materials), temporarily serving as a structured processing interface for expansion; teaching exercises (structured exercise questions and answers), serving as an auxiliary for misunderstanding pattern - knowledge point annotation.

[0154] S402, Construct a basic node and relationship network according to the core relationship pairs to obtain an initial graph.

[0155] Exemplarily, these core relationship pairs can be extracted from a table or database, organized into a preliminary data structure. For the four types of objects (misunderstanding patterns, knowledge points, teaching strategies, and teaching resources) in all core relationship pairs, each unique item is marked as a graph entity (such as a node ID), that is, the misunderstanding patterns, knowledge points, teaching strategies, and teaching resources in each core relationship pair are regarded as nodes in the initial graph.

[0156] Connect different nodes through the relationships between the four types of objects in the core relationship pairs to construct an initial graph. A graph database (such as Neo4j) can be used to store these nodes and relationships. For example, misunderstanding pattern - knowledge point: The misunderstanding pattern and the knowledge point are connected by a "causes" relationship, indicating that students may have an inertia misunderstanding when learning Newton's first law; knowledge point - teaching strategy: The knowledge point and the teaching strategy are connected by a "requires" relationship, indicating that this knowledge point needs to be explained through a strategy of explaining the concept of inertia; teaching strategy - teaching resource: The teaching strategy and the teaching resource are connected by a "uses" relationship, indicating that this teaching strategy can be implemented by playing a video of Newton's three laws.

[0157] S403, Use natural language processing technology to analyze teaching texts and extract the knowledge points of each text, and use computer vision technology to analyze teaching videos and extract the knowledge points of each video.

[0158] Exemplarily, candidate phrases in teaching texts can be extracted based on statistics (such as TF-IDF) or graph algorithms (such as TextRank), or context-related terms can be extracted based on deep models such as BERT; knowledge point definition structures (such as "X is...", "The properties of Y include...") can be identified using sentence templates; phrase nesting (such as the properties of the tangent of a circle) can be identified by applying syntactic analysis; the extracted terms are compared with a knowledge point thesaurus or a standard teaching knowledge system for deduplication or synonym merging (such as Newton's first law = law of inertia). Through these steps, a set of knowledge points for each text can be obtained. For example, for text ID: text_001, the knowledge points include Newton's first law and inertial reference frame.

[0159] An ASR (Automatic Speech Recognition) model can be used to convert the teacher's lecture speech into text (such as using DeepSpeech, Whisper, etc.), and the subsequent processing is the same as that for extracting knowledge points from teaching texts. Static key frames can be extracted using frame sampling or change detection to locate the moments of blackboard writing changes and PPT switches. Optical Character Recognition (OCR) is performed on the key frames to recognize the text, formulas, and PPT titles on the blackboard writing. The image text is matched with a dictionary to extract term-like phrases as candidate knowledge points. An image partitioning model (such as YOLO, LayoutLM) can be used to recognize the blackboard area, diagrams, arrow directions, etc., and perform image classification or object detection on visual elements (such as function images, physical schematic diagrams), and match the images with a knowledge point label library (such as a parabola trajectory diagram → projectile motion). The text knowledge points from ASR can be fused with the image knowledge points obtained from OCR / image understanding, that is, calculate the semantic similarity (such as BERT similarity) to determine whether they are the same knowledge point, and assign a confidence level (such as speech frequency + image position + visual clue strength), and use multi-modal voting to enhance the accuracy of knowledge point extraction. Through these steps, a set of knowledge points for each video segment can be obtained. For example, for video ID: video_007, the knowledge points include the kinetic energy theorem, the definition of work, and the velocity-time graph.

[0160] S404, fuse the text and video of the same knowledge point and connect them to the corresponding knowledge point in the initial knowledge graph to obtain the strategic knowledge graph.

[0161] It can be understood that although the teaching resources are already included in the core relationship pairs, the role of multi-modal data sources is not to repeatedly provide resources, but to supplement, verify, enhance, and dynamically update the relevance of teaching resources and the depth of knowledge understanding in the knowledge graph.

[0162] The teaching resources in the core relationship pairs are in a structured or annotated form (such as resource ID, resource name, or link), and do not include in-depth semantic analysis of the resource content. Multi-modal data sources provide content-level understanding of the resource content through semantic analysis, etc., enabling the knowledge graph to not only know that resource A is used for knowledge point X, but also know which knowledge points are specifically explained in resource A, how they are explained, and the depth of the explanation.

[0163] The core relationship pairs may come from expert input, template mining, or rule generation, and there are lags, limitations, or biases. Multi-modal data sources can be used to verify the fitness of existing strategies or resources with knowledge points, discover missing connections (such as knowledge point X has a video explanation but is not associated with the strategy), and complement potential teaching misunderstandings or alternative resources.

[0164] Exemplarily, the semantic similarity between the knowledge points of the teaching text and the knowledge points of the teaching video can be calculated, represented by vectors (such as BERT, word2vec embeddings), to determine whether they are the same or equivalent knowledge points. A similarity threshold is set for matching. The text and video identified as the same knowledge points are merged into the same node to form a set of fused nodes. The language context (definitions, example sentences) extracted from the text can be used as text attributes, and the image frames, audio segments, or image tags identified in the video can be used as visual attributes. In this way, each fused node becomes a structure with multi-modal features, which can be used for reasoning or recommendation.

[0165] Match the fused nodes with the knowledge point nodes in the initial knowledge graph (by knowledge point ID or standard name). If the knowledge point corresponding to the node already exists in the initial knowledge graph, the fused node is used as its supplementary information node or extended instance node, and semantic edges such as has_instance, described_by, or supported_by are established in the knowledge graph structure. If there is no knowledge point corresponding to the node in the graph, a new knowledge point node is automatically created and reconnected with relationships such as strategies and misunderstanding patterns. Edges can be established between the fused nodes and the knowledge point nodes in the initial knowledge graph. The edge types can include multi_modal_support (indicating that the knowledge point is jointly supported by multi-modal data), text_evidence (with text descriptions), and video_evidence (with video explanations). These edges may contain additional attributes such as evidence strength, timestamps, and contexts.

[0166] After the fused nodes are connected to the knowledge point entities in the initial knowledge graph, they will automatically inherit or be connected to all existing relationships of that knowledge point, thus obtaining a strategic knowledge graph. Each knowledge point entity in the strategic knowledge graph will have multi-modal evidence nodes as support, enhancing the interpretability and retrievability of teaching strategies. Strategy recommendation, path query, and resource location can be performed based on the knowledge graph. The strategy nodes in the graph can point back to the fused nodes that support them, forming a causal closed-loop path (such as misunderstanding pattern → knowledge point → teaching strategy → video resource).

[0167] This step ensures that the strategic knowledge graph not only has structural integrity but also integrates multi-source content understanding, making it have stronger reasoning ability, resource scheduling ability, and teaching strategy adaptation ability.

[0168] In a possible implementation, please refer to Figure 3 , the method for generating group teaching strategies based on the intelligent learning cloud platform further includes: S10, obtaining feedback data.

[0169] Exemplarily, after the generated group teaching strategy is implemented, feedback data of students can be obtained through the intelligent learning cloud platform, such as students' teaching behavior data and biological data after the implementation of the group teaching strategy.

[0170] S20. According to the feedback data, use the cognitive contradiction field to identify the feedback error pattern and calculate the feedback error intensity to obtain the first data. The first data includes the feedback error pattern and the feedback error intensity.

[0171] It can be understood that the cognitive contradiction field is an abstract model that simulates the cognitive energy state and can be used to represent the difference between the student's cognitive state and the target state, mapping the student's current learning state to a multi-dimensional cognitive space. If a student has a misunderstanding about a certain knowledge point, a cognitive energy trap, also known as a cognitive conflict area, will be formed at this node, indicating that there is a deviation between the student's current cognitive state and the target cognitive state.

[0172] Exemplarily, through the feedback data of students, a comparison graph between the student's answering or learning path and the ideal learning path can be constructed to identify at which steps or subtasks the deviation occurs, and use the path difference to locate the cognitive obstacle points (i.e., the occurrence points of the error pattern). For example, a student repeatedly uses the wrong formula when answering questions involving Newton's second law. It can be judged from the answering process and historical records that the student has a misunderstanding of misusing the formula; if a student frequently confuses the difference between kinetic energy and potential energy, it can be determined that there is a concept confusion; if a student always skips intermediate steps, it can be identified as a broken reasoning chain or missing steps. The identified error type is the feedback error pattern, which is an error manifestation with cognitive deviation characteristics.

[0173] After identifying the feedback error pattern, the frequency of the error occurring in the group can be calculated. If a certain misunderstanding appears repeatedly in multiple tests or assignments, it indicates that the problem is deeply rooted and the error intensity is higher. Some knowledge points are at key positions in the learning path. For example, the quadratic function graph is the basis for subsequent analytic geometry problems. If a misunderstanding occurs at an important node, the intensity score of this error can be increased. If an error has repeatedly occurred in multiple past learning cycles of the group, this error is determined to be a persistent misunderstanding and the intensity is further increased. It can also be analyzed whether this error may affect other knowledge points. If an error may cause students to make mistakes in multiple related knowledge points (such as confusing speed and acceleration will affect multiple parts of mechanics), its intensity will be further increased. Finally, the above dimensions are weighted and summarized to calculate a numerical error intensity to represent the depth of influence of this misunderstanding in the student's cognitive structure.

[0174] This process forms a mapping from feedback data to error representations in the cognitive sense, enabling cognitive problems to be introduced into the strategy optimization logic in an identifiable, quantifiable, and intervenable manner, which is the key cognitive basis for strategy optimization.

[0175] S30. Based on the first data, filter out multimodal data from the policy knowledge graph, and extract the association between the feedback error pattern and the multimodal data to obtain the second data. The second data includes the multimodal data and the association between the feedback error pattern and the multimodal data.

[0176] Exemplarily, use the feedback error pattern in the first data as the key query condition, and combine the associated knowledge points of the feedback error pattern to search for the connected policy path in the policy knowledge graph. Through traversing the policy knowledge graph and semantic matching, it can be identified which teaching resources are related to this error and which resources have been bound to intervene in this type of error.

[0177] Among the identified resource nodes, video clips strongly related to the target knowledge point / error pattern can be filtered out according to the modal type, obtained through automatic speech recognition (ASR) and image tagging in the video; paragraphs semantically matching the error pattern extracted from the teaching text. Augmented reality scenarios interactively bound to the knowledge point (error pattern) can also be filtered out, such as simulating the kinetic energy conversion process; practice questions related to the error can be filtered out for verification or guiding correction. The attributes of each resource node can include the modal type, knowledge point coverage, past intervention effect indicators (such as the error correction ratio of this resource in the group), and semantic similarity score (the semantic fit of the resource content with the feedback misunderstanding).

[0178] Based on the structure of the policy knowledge graph, the association path between each resource and the error pattern can be extracted to identify whether the resource has been used to directly intervene in this error pattern, indirectly support the knowledge points associated with this misunderstanding, or whether the knowledge misunderstanding is clearly explained or demonstrated in its content.

[0179] Natural language processing and visual semantic embedding techniques can be used to further calculate the matching degree with the misunderstanding pattern from the resource content. For example, whether the speech of the video explanation contains common misunderstandings in formula selection, whether the text presents in the form of wrong examples, and whether the AR scenario simulates the process of misunderstanding correction through interaction. An association strength score can be assigned to each pair of error pattern - multimodal resource relationships for subsequent sorting and policy reasoning.

[0180] Through the above steps, the semantic-level docking of the structured modeling results of cognitive errors and the teaching resources in the policy knowledge graph is completed, providing high-quality and multi-dimensional input support for the subsequent dynamic optimization of teaching strategies.

[0181] S40. Based on the feedback data, the first data, and the second data, use the reinforcement learning algorithm to update the policy knowledge graph.

[0182] Exemplarily, the core elements of reinforcement learning include State, Action, and Reward. The state includes the execution state of each policy in a specific teaching scenario; the current knowledge point, error pattern, the applied policy and its resource path. For example, on knowledge point K1, for error pattern M1, use policy S1 + resource V1. Actions include policy operations that can be taken on the current state, such as retaining the current policy, replacing it with other policies, adding supplementary resources, adjusting the priority or scope of the policy. Rewards include quantitative metrics driven by feedback data, used to judge the effectiveness of the policy. For example, a decrease in error intensity → positive reward, an improvement in academic performance → positive reward, ineffective multimodal resource intervention → negative reward, no obvious improvement in learning time → weak reward or punishment.

[0183] Based on the student feedback data generated after the execution of each teaching policy, it can be analyzed whether the student has made cognitive progress under the intervention of this policy; it can be judged whether the policy is effective based on whether the error intensity in the first data has decreased; combined with the multimodal resource relevance in the second data, it can be judged whether the resource has played the expected role.

[0184] According to the feedback effect, assign a reward value to each policy path. The reward value is not only based on the current result but can also incorporate historical performance (such as using a sliding window). The reward mechanism can include a significant reduction in error intensity, highly effective policy, +1 to the reward value; neutral effect, no significant change, reward value remains unchanged; misunderstanding persists or worsens, the policy may be ineffective or have side effects, -1 to the reward value.

[0185] Based on the cumulative reward value and the current policy effect, the credibility, priority weight of the current policy in the policy knowledge graph can be increased, or the activation probability of the current policy can be reduced, and the applicable scope can be narrowed; other policy paths related to this error can be introduced into the policy knowledge graph (possibly migrated from historical data, the expert library, or other nodes); multiple sub - policies can be combined into a new composite policy to deal with complex misunderstandings.

[0186] Based on the output results of the reinforcement learning policy, structural and property updates can be made to the policy knowledge graph. The attributes of policy nodes can be modified, such as policy trust, applicable misunderstanding pattern range, resource adaptability score, reinforcement weight; the edges between the policy and the knowledge point, misunderstanding pattern, multimodal resource can be strengthened or weakened, that is, increase the weight of the edge, indicating a stronger association, close / delete the edges with poor effects, indicating resource failure, create new edges, indicating that new policies or new resources are found to be available for intervening in this error; identify new effective policy paths in the graph and incorporate them into the recommended policy library for subsequent policy recommendation or autonomous generation.

[0187] The reinforcement learning process is continuously executed in a loop. After each round of policy execution, new feedback data is generated and enters the next learning cycle, gradually realizing the personalization of policy selection, the optimization and compression of policy paths, the precise matching of multi-modal resources and error types, and the dynamic evolution of the graph structure, making it more in line with the cognitive changes of students in real teaching scenarios.

[0188] Through the above steps, the system not only has the ability of self-learning, but also can continuously improve the policy recommendation according to the real teaching results, which is beneficial to the adaptability and dynamic optimization of teaching strategies for different student groups, and realizes a high-quality and sustainable intelligent teaching support system.

[0189] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0190] Corresponding to the method for generating group teaching strategies based on the intelligent learning cloud platform described in the above embodiments, an embodiment of the present application also provides a device for generating group teaching strategies based on the intelligent learning cloud platform. Each unit of the device can implement each step of the method for generating group teaching strategies based on the intelligent learning cloud platform.

[0191] The device includes: An acquisition unit, configured to acquire teaching behavior data, biological data, and teaching progress data of multiple classes through the intelligent learning cloud platform. Among them, the teaching progress data includes the taught knowledge points and the corresponding teaching dates.

[0192] A spatio-temporal alignment unit, configured to perform spatio-temporal alignment processing on the teaching stages of multiple classes according to the teaching progress data of multiple classes to obtain a teaching stage sequence. Among them, the teaching stage sequence includes multiple teaching stages, and each teaching stage includes the teaching behavior data and biological data of multiple classes in the same teaching stage.

[0193] A group anomaly detection unit, configured to perform group anomaly detection on multiple classes based on each teaching stage of the teaching stage sequence to obtain an anomaly detection result. Among them, the anomaly detection result includes the teaching stages with group anomalies.

[0194] A group teaching strategy generation unit, configured to generate a group teaching strategy based on the teaching behavior data and biological data of the teaching stages in the anomaly detection result.

[0195] It should be noted that the information interaction, execution process, etc. between the above units, due to being based on the same concept as the method embodiment of the present application, for their specific functions and the technical effects brought, reference can be specifically made to the method embodiment part, and details are not described here again.

[0196] The embodiments of the present application further provide an electronic device. The electronic device in this embodiment includes: at least one processor, at least one memory, and a computer program stored in the at least one memory and executable on the at least one processor. When the processor executes the computer program, the electronic device implements the steps in any of the above-described embodiments of the group teaching strategy generation method based on the intelligent learning cloud platform, or the functions of each unit in the above-described device embodiments.

[0197] Exemplarily, the computer program may be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present application. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and these instruction segments are used to describe the execution process of the computer program in the electronic device.

[0198] The electronic device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The electronic device may include, but is not limited to, a processor and a memory. It may include more or fewer components, or combine certain components, or different components. For example, it may also include input / output devices, network access devices, a bus, etc.

[0199] The processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0200] The memory may be an internal storage unit of the electronic device in some embodiments, such as a hard disk or memory of the electronic device. The memory may also be an external storage device of the electronic device in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device. Further, the memory may also include both the internal storage unit and the external storage device of the electronic device. The memory is used to store an operating system, application programs, a BootLoader, data, and other programs, such as program codes of the computer program. The memory may also be used to temporarily store data that has been output or is to be output.

[0201] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For parts not detailed or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0202] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included within the protection scope of the present application.

Claims

1. A method for generating a group teaching strategy based on a smart learning cloud platform, characterized in that, Including: Obtaining teaching behavior data, biological data, and teaching progress data of multiple classes through a smart learning cloud platform; wherein, the teaching progress data includes the taught knowledge points and the corresponding teaching dates; According to the teaching progress data of multiple classes, performing spatio-temporal alignment processing on the teaching stages of multiple classes to obtain a teaching stage sequence; wherein, the teaching stage sequence includes multiple teaching stages, and each teaching stage includes the teaching behavior data and biological data of multiple classes in the same teaching stage; Based on each teaching stage of the teaching stage sequence, performing group anomaly detection on multiple classes to obtain an anomaly detection result; wherein, the anomaly detection result includes the teaching stages with group anomalies; Generating a group teaching strategy based on the teaching behavior data and biological data of the teaching stages in the anomaly detection result.

2. The method for generating a group teaching strategy based on an intelligent learning cloud platform according to claim 1, wherein, The performing spatio-temporal alignment processing on the teaching stages of multiple classes according to the teaching progress data of multiple classes to obtain a teaching stage sequence includes: Comparing whether the knowledge points in the teaching progress data of multiple classes are consistent with the knowledge points in the standard teaching framework. If they are inconsistent, according to the equivalent knowledge point table, replacing the inconsistent knowledge points in the teaching progress data with the corresponding knowledge points in the standard teaching framework; wherein, the equivalent knowledge point table includes the equivalent relationships between the same or similar knowledge points in different textbooks and the knowledge points in the standard teaching framework, and the equivalent knowledge point table is obtained through semantic similarity calculation and manual rule base supplementation. The standard teaching framework includes each teaching stage ID, the knowledge points included in each teaching stage, and the expected teaching duration of each teaching stage; Calculating the distance matrix between the knowledge points in the teaching progress data of each class and the knowledge points in the standard teaching framework; based on each distance matrix, using the dynamic time warping algorithm to calculate the optimal alignment path between the teaching progress of each class and the standard teaching framework; according to the optimal alignment path between the teaching progress of each class and the standard teaching framework, mapping the teaching progress of each class to each teaching stage in the standard teaching framework to obtain a teaching stage sequence.

3. The method for generating a group teaching strategy based on an intelligent learning cloud platform according to claim 2, wherein The method further includes: Extracting knowledge points from the teaching content based on text analysis technology; wherein, the teaching content is pre-stored in the smart learning cloud platform; Using a relationship extraction algorithm to analyze the context in the teaching content, identifying the dependency relationships between the knowledge points, and constructing a knowledge point dependency graph; According to the knowledge point dependency graph, dividing the knowledge points into each teaching stage and labeling each teaching stage with an ID; Allocating the expected teaching duration of each teaching stage based on the number of knowledge points in each teaching stage; Arranging each teaching stage ID, the knowledge points included in each teaching stage, and the expected teaching duration of each teaching stage according to the dependency relationships between the knowledge points to obtain the standard teaching framework.

4. The method for generating a group teaching strategy based on an intelligent learning cloud platform according to claim 2, wherein The method further includes: During the spatio-temporal alignment process, use the dynamic time warping algorithm to detect whether there is a progress jump feature; wherein, the progress jump feature means that the knowledge points corresponding to the teaching progress of a certain class in the standard teaching framework are not covered by the teaching progress data of this class; If there is the progress jump feature, insert virtual knowledge points at the positions of the knowledge points not covered in the teaching stage sequence to fill the knowledge points not covered; Or If there is the progress jump feature, mark the knowledge points not covered in the teaching stage sequence and indicate the absence of the knowledge points not covered.

5. The method for generating a group teaching strategy based on a smart learning cloud platform according to claim 1, characterized in that, Based on each teaching stage of the teaching stage sequence, perform group anomaly detection on multiple classes to obtain an anomaly detection result, including: Extract the teaching behavior data and biological data of multiple classes in each teaching stage from the teaching stage sequence; Perform the following processing for each teaching stage: Extract the common error patterns from the teaching behavior data of multiple classes; construct a cognitive network graph based on the common error patterns, and based on the cognitive network graph, identify group knowledge breakpoints; integrate the common error patterns and the group knowledge breakpoints to obtain group cognitive features; wherein, the cognitive network graph is used to represent the mastery situation and correlation of different knowledge points by students during the learning process; Based on the teaching behavior data of multiple classes, use Fourier transform to perform frequency analysis on the learning behavior of students, identify high-frequency and low-frequency components to obtain a behavior frequency spectrum; calculate the resonance matching degree between the behavior frequency spectrum and the ideal learning path frequency spectrum; based on the resonance matching degree, identify non-explicit behavior patterns to obtain group behavior features; Calculate the stress index according to the biological data of multiple classes to obtain group emotion features; Perform feature fusion on the group cognitive features, the group behavior features and the group emotion features to obtain fusion features; Compare the fusion features with the dynamic anomaly threshold. If the fusion features are greater than the dynamic anomaly threshold, determine that there is a group anomaly in this teaching stage to obtain the anomaly detection result.

6. The method for generating a group teaching strategy based on a smart learning cloud platform according to claim 5, wherein The method further includes: Obtain the teaching behavior data of all classes in multiple historical teaching stages, and determine the teaching behavior data of all classes in each historical teaching stage as a meta-task; For each meta-task, divide the teaching behavior data of all classes into a training set and a test set; train the meta-learning model based on the training set, and calculate the loss value of the meta-learning model on the training set; according to the loss value, update the model parameters of the meta-learning model through the gradient descent algorithm; calculate the test loss based on the test set, and calculate the gradient of the test loss with respect to the model parameters based on the test loss; Calculate the average value of the gradients of all meta-tasks to obtain a meta-gradient; update the model parameters of the meta-learning model according to the meta-gradient to obtain the trained meta-learning model; Input the teaching behavior data of multiple classes in each teaching stage into the meta-learning model after training, so that the meta-learning model after training can extract the behavior gradient and balance coefficient of each teaching stage, and determine the behavior gradient and balance coefficient of each teaching stage as the group behavior characteristics of each teaching stage.

7. The method for generating a group teaching strategy based on an intelligent learning cloud platform according to claim 5, wherein The method further includes: Calculate the mean and standard deviation of the historical group data, and determine the sum of twice the mean and standard deviation of the historical group data as the initial anomaly threshold; wherein, the historical group data includes the fusion features of each teaching stage in the teaching stage sequences of multiple historical cycles. Based on the fusion features of all teaching stages in the teaching stage sequence, calculate the standard deviation of the teaching stage sequence. Compare the standard deviation of the historical group data with the standard deviation of the teaching stage sequence to obtain a comparison result, and adjust the multiple of the standard deviation in the initial anomaly threshold according to the comparison result to obtain the dynamic anomaly threshold.

8. The method for generating a group teaching strategy based on an intelligent learning cloud platform according to claim 5, wherein, Generating a group teaching strategy based on the teaching behavior data and biological data of the teaching stage in the anomaly detection result includes: Select the associated strategies matching the group cognitive characteristics from the strategy knowledge graph; perform associated expansion on the common error patterns in the group cognitive characteristics to obtain extended features, and retrieve the extended strategies matching the extended features from the strategy knowledge graph; integrate the associated strategies and the extended strategies to obtain an initial teaching strategy. Based on the group behavior characteristics, adjust the strategy form of the initial teaching strategy, and based on the group emotion characteristics, adjust the strategy attributes of the initial teaching strategy to generate a group teaching strategy.

9. The method for generating a group teaching strategy based on an intelligent learning cloud platform according to claim 8, wherein The method further includes: Obtain core relation pairs and multimodal data sources; wherein, the core relation pairs include the relationships between misunderstanding patterns, knowledge points, teaching strategies, and teaching resources, and the multimodal data sources include teaching texts and teaching videos. Construct a basic node and relationship network according to the core relation pairs to obtain an initial graph. Use natural language processing technology to analyze the teaching text and extract the knowledge points of each text in the teaching text, and use computer vision technology to parse the teaching video and extract the knowledge points of each video in the teaching video. Fuse the texts and videos with the same knowledge points and connect them to the corresponding knowledge points in the initial graph to obtain the strategy knowledge graph.

10. The method for generating a group teaching strategy based on an intelligent learning cloud platform according to claim 9, wherein, The method further includes: Obtain feedback data. According to the feedback data, use the cognitive contradiction field to identify the feedback error pattern and calculate the feedback error intensity to obtain the first data; wherein, the first data includes the feedback error pattern and the feedback error intensity. According to the first data, screen out multimodal data from the strategy knowledge graph and extract the association between the feedback error pattern and the multimodal data to obtain the second data; wherein, the second data includes the multimodal data and the association between the feedback error pattern and the multimodal data. Based on the feedback data, the first data, and the second data, use the reinforcement learning algorithm to update the strategy knowledge graph.

Images