Information-based teaching test system with adaptive learning progress

Abstract

The application relates to the technical field of intelligent education management, in particular to an information-based teaching test system with adaptive learning progress, which comprises the following contents: a knowledge node construction module, a learning state recognition module, a test task generation module, a path scheduling adjustment module and a feedback data backtracking module.In the application, the knowledge point fine numbering management is realized through the construction of a directed edge numbering set, the dependence direction is set in combination with a label propagation algorithm, the semantic expression capability of the node relationship is enhanced, the state weight sequence is formed based on the matching of the answer behavior matching weight and the stability coefficient, the stability of the classification recognition is improved, the cognitive ability coding and the time consumption median value linkage state weight are introduced, the questions are filtered and a priority queue is constructed, the precision of the task pushing is improved, the path chain is constructed by fusing the numbering and the sorting weight, the path branch is expanded by calculating the dependence strength, and the individual adaptability and the task rhythm coordination capability of the path scheduling are enhanced.

Description

Technical Field

[0001] This invention relates to the field of intelligent education management technology, and in particular to an information-based teaching and testing system that adapts to learning progress. Background Technology

[0002] The field of intelligent education management technology encompasses the construction of educational information systems, AI-based optimization of teaching processes, student learning behavior analysis, digital processing of teaching resources, and intelligent evaluation mechanisms. Its core lies in data-driven management of the entire teaching process through computer systems. This includes comprehensive integration of aspects such as teaching resource distribution, learning progress monitoring, student behavior data collection and processing, and teaching feedback generation. Systematically, the development of intelligent education management technology covers learning behavior perception, data modeling, learning path planning, and intelligent recommendation mechanisms. Through continuous analysis and dynamic response to individual student learning trajectories, it constructs a learner-centered information-based teaching environment, thereby supporting multi-dimensional and multi-level educational management and strategic decision-making.

[0003] Among them, the information-based teaching and testing system for adaptive learning progress refers to a management system that adjusts the pace of subsequent learning tasks in real time based on factors such as learners' historical learning behavior data, current task completion status, and learning ability assessment indicators. The technical aspects it addresses include learning task duration prediction, individual difference identification, learning priority judgment, and pace adjustment decision-making. It constructs a feature matrix by setting multi-dimensional input variables, calls a trained clustering model to divide learner groups, matches corresponding learning resource paths based on a rule engine, and combines a segmented time window task monitoring mechanism to determine the individual's learning progress rate and the adjustment logic of test content, thereby dynamically pushing the most suitable test nodes and question sequences.

[0004] Existing technologies often rely on static course structures and fixed label relationships to generate knowledge point networks, lacking semantically driven structural derivation mechanisms. This results in insufficient expression of hierarchical connections between knowledge points. Answer behavior analysis often focuses on accuracy statistics and coarse-grained behavioral characteristics, neglecting the mining of deeper features such as stability and cognitive fluctuations, affecting the comprehensiveness and accuracy of mastery judgment. Task generation strategies are usually based on rule matching and question bank screening, failing to fully integrate the dynamic features in behavioral data, resulting in a lack of precision control and real-time adaptation capabilities for task delivery. The path scheduling process often follows a linear progress, failing to effectively identify the dependence strength between tasks and the impact of individual differences on path construction, easily leading to unbalanced task rhythm or repetitive path arrangements, affecting learning motivation and efficiency. In the teaching and testing process, due to the failure to identify structural differences in learners' mastery levels across multiple knowledge points, problems such as repetitive test content, excessive jumps, or unbalanced difficulty often occur, affecting the assessment effect and the scientific nature of subsequent learning arrangements. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide an information-based teaching and testing system with adaptive learning progress. The technical solution is as follows:

[0006] On the one hand, it provides an information-based teaching and testing system that adapts to the learning pace. This system includes:

[0007] The knowledge node construction module extracts the course knowledge structure to obtain knowledge point nodes, constructs directed edges and generates a set of numbers, extracts teaching tags and course outline structure, uses the tag propagation algorithm to set the direction of dependency edges, constructs a directed graph structure, generates a set of knowledge point numbers and passes it to the learning status recognition module.

[0008] The learning state recognition module calls the knowledge point number set, obtains the answering behavior, matches the mastery weight and stability coefficient, records the state weight, uses the support vector machine to classify the sequence, identifies the mastery state, generates a learning state node set and passes it to the test task generation module.

[0009] The test task generation module calls the learning state node set, filters the mastery state nodes, extracts the cognitive ability code and median time of the test questions, combines the state weight and the change in answering time to filter questions, constructs a priority queue for push tasks and passes it to the path scheduling and adjustment module.

[0010] The path scheduling and adjustment module calls the push task priority queue, constructs a learning path chain based on the knowledge point number, extracts the path number and sorting weight, calculates the node dependency strength score, expands the path chain to form a path branch set, generates the path expansion range and passes it to the feedback data backtracking module.

[0011] As a further embodiment of the present invention, the knowledge point number set includes knowledge point encoding, dependency edge attributes, and node graph sequence identifier; the learning state node set includes mastery state label, state confidence score, and answer performance characteristics; the push task priority queue includes test question number, matching scoring threshold, and scheduling order index; and the path extension interval includes path number sequence, branch node group, and extended scoring index.

[0012] As a further aspect of the present invention, the knowledge node construction module includes:

[0013] The structure parsing submodule extracts the course knowledge structure to obtain knowledge point nodes, calls the course teaching tags and course outline structure content, identifies the tag field corresponding to each node, classifies and encodes the nodes in combination with the course structure hierarchy, identifies the topic category and index position, and establishes node distribution hierarchy values.

[0014] The directed edge construction submodule calls the node distribution hierarchy value, identifies direct reference pairs between nodes based on the category and order relationship of the nodes in the structure, establishes a set of directed connections based on the reference direction and structural hierarchy, and uses a label propagation algorithm based on the label weight allocation mechanism of the course structure hierarchy to calculate the label dependency degree for each node pair, determine the propagation direction of the edge, and generate a set of directed dependency paths.

[0015] The number set generation submodule extracts the node numbers and sorting positions in the dependent paths based on the directed dependency path set, adjusts the numbering order according to the order relationship of the nodes in the path, combines the node position index, connection direction and path sequence to obtain the number identifier, and establishes the knowledge point number set.

[0016] As a further aspect of the present invention, the learning state recognition module includes:

[0017] The behavior extraction submodule calls the knowledge point number set to obtain the answer time, score and number of modifications in each question's answer record of the learner, performs data item standardization transformation on each type of data, classifies the standardization results in combination with the knowledge point number, and generates answer behavior association values.

[0018] The state matching submodule calls the answer behavior association value, compares and analyzes each data item by item according to the set mastery weight benchmark value and stability coefficient benchmark value, analyzes the mastery matching status of the behavior corresponding to each knowledge point, records the node state offset direction and offset magnitude, and generates a state offset coefficient set.

[0019] The mastery classification submodule extracts the mastery weight value, stability coefficient, and offset amplitude value corresponding to each node based on the state offset coefficient set. By extracting the fluctuation rate of answering time, score deviation, accuracy rate, and progress rate, and using support vector machine, it identifies the learner's mastery status on each knowledge point, including mastery, partial mastery, and no mastery, establishes a classification structure table, and obtains the learning state node set.

[0020] As a further aspect of the present invention, the test task generation module includes:

[0021] The master node filtering submodule calls the learning state node set, identifies nodes labeled as mastered, extracts the corresponding node number and state label value, and generates a master node number set.

[0022] The question ability extraction submodule calls the mastery node number set, extracts the test question information corresponding to the node, extracts the cognitive ability code and standard time data of the test question, calculates the median time value of the same type of question within the node range, statistically analyzes the time fluctuation range and ability code coverage level of the question, and obtains the cognitive feature value of the test question.

[0023] The priority queue construction submodule calculates the median time consumption and state weight difference for each question based on the cognitive feature values ​​of the test questions, calculates the task sorting priority based on the state deviation and cognitive encoding complexity, filters test questions, and establishes a priority queue for push tasks.

[0024] As a further aspect of the present invention, the specific formula for calculating the task ranking priority is as follows:

[0025] ;

[0026] Calculate task sorting priority;

[0027] in, Representing the Prioritization of test questions Representing the Median time taken for each test question Representing the The state weights of the test questions This represents the average of the state weights of all test items. Representing the Test questions and the The difference in cognitive ability encoding between knowledge points The average value representing the cognitive ability codes for all test items. Represents the total number of knowledge points. Represents the currently calculated number of The identifier of each knowledge point This represents the number of each individual test question in the system.

[0028] As a further aspect of the present invention, the path scheduling adjustment module includes:

[0029] The path construction submodule calls the push task priority queue, extracts the knowledge point numbers included in the task, sorts the number set in order, establishes the path connection relationship between knowledge points, configures the path identification number of the connection relationship, extracts the node sorting weight value in combination with the task order, and generates the learning path structure number set.

[0030] The dependency scoring submodule calls the learning path structure number set, identifies the connection edge information of adjacent nodes in the path chain, calculates the dependency weight value between each pair of nodes according to the sorting weight, and obtains the node dependency strength score value.

[0031] The path expansion submodule selects the two preceding nodes with the highest scores based on the node dependency strength score and uses them as the expansion benchmark. It then extracts the corresponding path number and expands the path chain, establishes the expanded path branch set, and generates the path expansion interval.

[0032] As a further aspect of the present invention, the specific formula for identifying the connection edge information of adjacent nodes in the path chain is as follows:

[0033] ;

[0034] Calculate the eigenvalues ​​of the complexity of connecting edges;

[0035] in, Representative node With nodes The complexity eigenvalues ​​of the edges connecting them. Represents the time slice in the path structure internal nodes With nodes Information strength value of the connecting edges This is the structural perturbation adjustment factor. Represents a node With nodes In time slice The difference in direction between inner connected edges. This is the direction weight correction factor. This is the minimum disturbance offset. This represents the total number of time slices counted in the path chain. Represents a node With nodes The edge distribution density value between them, Represents a node The mean of the edge distribution density values ​​between it and all its connected nodes. The index variable is the time slice number to be summed. This is the index number of the starting node in the path chain. This is the index number of the target node in the path chain.

[0036] As a further aspect of the present invention, the system further includes:

[0037] The feedback data backtracking module calls the task pre-path expansion interval to obtain the path node answer sequence and stability parameters, compares the accuracy and time consumption changes, analyzes the learning state change trend, and generates an update path structure task marking instruction.

[0038] The updated path structure task marking instruction includes node update marking, trend identification type, and feedback update parameters.

[0039] As a further aspect of the present invention, the feedback data backtracking module includes:

[0040] The answer data extraction submodule extracts the path number and node set corresponding to the extended path based on the path extension interval, collects the answer sequence data and stability coefficient corresponding to each node, performs node number classification processing on the answer data, generates node answer accuracy and time consumption records, and establishes node answer change values.

[0041] The state trend calculation submodule calls the node's response change value, performs a horizontal comparison of the node's accuracy and time consumption within the path, judges and grasps the state change trend based on the change direction, and calculates the trend score value in combination with stability parameters to obtain the learning state change trend value.

[0042] The path marker generation submodule extracts the node numbers whose trend score values ​​exceed the state adjustment threshold based on the learning state change trend value, matches the corresponding path number information, performs structural mapping and integration of the extracted nodes and paths, generates a path update identifier structure including node number, path number and adjustment type, and establishes a path structure update instruction set.

[0043] The state adjustment threshold is calculated by outlier score based on statistical features within a sliding window. If the score exceeds the set threshold, the learning state adjustment of the path structure is triggered.

[0044] The beneficial effects of the technical solutions provided by the embodiments of the present invention include at least the following:

[0045] By extracting knowledge point nodes from the course knowledge structure and introducing them into teaching tags and the semantic structure of the course syllabus, directed edges are constructed and uniformly numbered, effectively realizing the structured expression of knowledge points. A tag propagation algorithm is introduced to set the direction of dependency relationships, enabling the hierarchical logic between knowledge points to have semantic transmission capabilities, providing high-quality structural support for subsequent state recognition and path construction. Combining the mastery weight and stability coefficient during the question-answering behavior acquisition process, a state weight sequence oriented towards knowledge points is established, and the support vector machine algorithm is used for classification and recognition, significantly improving the accuracy of mastery state discrimination. Through the cognitive ability encoding and time consumption data analysis of learning state nodes, learner behavioral characteristics are accurately characterized, achieving dynamic adaptation in question selection. A dual-factor screening mechanism of state weight and answering time change is introduced in the task push stage, constructing a push priority queue to make test tasks more targeted and progressive. A path chain is constructed based on the number set, and a node dependency strength scoring mechanism is introduced to expand the generated path branch set, simultaneously realizing personalized process control and task difficulty and rhythm adjustment in path scheduling. The multi-stage data extraction and correlation fusion constructs a complete closed-loop process, which solves the problems of structural fragmentation, response delay and low adaptability in traditional learning path planning, and enhances the system's ability to respond to individual differences among learners and the dynamic adaptability of path recommendation. Attached Figure Description

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

[0047] Figure 1 This is a system flowchart of the present invention;

[0048] Figure 2 This is a schematic diagram of the system framework of the present invention; Detailed Implementation

[0049] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0050] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0051] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0052] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0053] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0054] This invention provides an information-based teaching and testing system that adapts to learning progress. Please refer to [link / reference]. Figures 1 to 2 This invention provides a technical solution: an information-based teaching and testing system that adapts to learning progress includes:

[0055] The knowledge node construction module extracts the course knowledge structure to obtain knowledge point nodes, constructs directed edges and generates a set of numbers, extracts teaching tags and course outline structure, uses the tag propagation algorithm to set the direction of dependency edges, constructs a directed graph structure, generates a set of knowledge point numbers and passes it to the learning status recognition module.

[0056] The learning state recognition module calls the knowledge point number set, obtains the answering behavior, matches the mastery weight and stability coefficient, records the state weight, uses support vector machine to classify the sequence, identifies the mastery state, generates a learning state node set and passes it to the test task generation module.

[0057] The test task generation module calls the learning status node set, filters the mastery status nodes, extracts the cognitive ability code and median time of the test questions, combines the status weight and the change in answering time to filter questions, builds a priority queue for push tasks and passes it to the path scheduling and adjustment module.

[0058] The path scheduling and adjustment module calls the priority queue of push tasks, constructs a learning path chain based on the knowledge point number, extracts the path number and sorting weight, calculates the node dependency strength score, expands the path chain to form a set of path branches, generates the path expansion range and passes it to the feedback data backtracking module.

[0059] The feedback data backtracking module calls the task pre-path extension interval, obtains the path node answer sequence and stability parameters, compares the accuracy and time consumption changes, analyzes the learning state change trend, and generates an instruction to update the path structure task marking.

[0060] The knowledge point number set includes knowledge point codes, dependency edge attributes, and node graph sequence identifiers; the learning status node set includes mastery status labels, status confidence scores, and answer performance characteristics; the push task priority queue includes test question numbers, matching scoring thresholds, and scheduling order indexes; the path expansion interval includes path number sequences, branch node groups, and expanded scoring indicators; and the path structure update task marking instructions include node update markings, trend identification types, and feedback update parameters.

[0061] Please see Figure 2 The knowledge node construction module includes:

[0062] The structure parsing submodule extracts the course knowledge structure to obtain knowledge point nodes, calls the course teaching tags and course outline structure content, identifies the tag field corresponding to each node, classifies and encodes the nodes in combination with the course structure hierarchy, identifies the topic category and index position, and establishes node distribution hierarchy values.

[0063] The structural analysis sub-module starts with the extraction of the curriculum knowledge structure, calls the teaching tags and the content of the curriculum syllabus structure. In the specific implementation process, first, it is necessary to extract the title information, sub-titles and key terms of each chapter in the curriculum content, and compare and match them with the module structure listed in the teaching syllabus. For example, in a "Higher Mathematics" course, "Derivatives and Differentials", "Definite Integrals", and "Function Limits" are used as the first-level structure nodes respectively. At the same time, second-level tags such as "Definition", "Geometric Meaning", and "Calculation Rules" are extracted from the syllabus for corresponding mapping. Then, these nodes are classified and coded according to the structural hierarchy. The classification coding needs to be numbered based on the chapter level of the node in the curriculum. For example, the first-level node coding is set as A, the second-level node is A1, A2, and the third-level node is A1.1, A1.2. During the classification process, teaching tags such as "Basic Concepts", "Example Analysis", and "Knowledge Expansion" are called, and the structural hierarchy position is judged according to the order in which they appear in the syllabus. For example, if "Basic Concepts" is at the beginning of the chapter, its classification level is 1. Further, by analyzing the topic relationships between different tags and nodes in the curriculum, using the method of attributing topic keywords, it is identified that the nodes belong to different topic categories such as "Calculus" and "Linear Algebra", and index positions are assigned according to their order in the curriculum system. Suppose the "Derivatives" chapter is taught in the 3rd week, its index position is 3. Combining the above classification coding and position index, the distribution hierarchy value of each node is finally constructed. For example, the A1.2 node represents the second sub-point of the first section of Chapter A, and its hierarchy value is (1,2). In this process, the identification of the tag field adopts the method of pure text retrieval and comparison. Suppose the tag set is , for each curriculum node , match the corresponding fields in it to generate a tag set , and then use the matching position and the order relationship of the tags in the syllabus to determine its classification number. When performing the "judgment" action, compare the level number i of the node with its tag index j. When i < j, it is marked as a superior node, and vice versa as an inferior node. Suppose the reference level is 2. If the node "Mean Value Theorem for Differentials" has the number 3.1.2 and its tag position is at the 2nd level, then it is judged as an inferior node. Here, the reference level 2 is obtained through curriculum structure statistics. Suppose the curriculum has a total of 5 levels of structure, then the middle value 3 is selected as the reference level, and the following interval is divided: level values 1 - 2 are upper-level nodes, 3 is the middle-level node, and 4 - 5 are lower-level nodes. Finally, a distribution hierarchy table of the nodes is established, as shown in Table 1;

[0064] Table 1 Distribution Hierarchy Table of Curriculum Nodes

[0065]

[0066] As shown in Table 1, the nodes are systematically distributed according to the course structure hierarchy. Based on the label type and the position in the course outline, the hierarchical value and index position of each node are obtained to ensure that the node classification coding has a hierarchical logical order.

[0067] The directed edge construction submodule calls the node distribution hierarchy value, identifies direct reference pairs between nodes based on the category and order of nodes in the structure, establishes a set of directed connections based on the reference direction and structural hierarchy, and uses a label propagation algorithm based on the label weight allocation mechanism of the course structure hierarchy to calculate the label dependency degree for each node pair, determine the propagation direction of the edge, and generate a set of directed dependency paths.

[0068] The directed edge construction submodule is based on the reference relationships between nodes. It calls the node distribution level values. During execution, it first extracts the level value combination of each pair of adjacent nodes. For example, for nodes A1.1 and A1.2, the level values ​​are 1 and 2 respectively. Directed connecting edges are then established according to their order in the course structure.<A1.1→A1.2> Next, all node pairs with consecutive hierarchical values ​​but different label affiliations are identified as direct reference pairs. For example, a "Basic Concepts" label node immediately followed by an "Example Analysis" node can be identified as a reference example for that knowledge point. When constructing edge pairs, a directional relationship is used to form a unidirectional link from lower to higher levels. In the "judgment" operation, the label dependency degree of the formed edges is calculated based on the label type. Let the weight of the label "Basic Concepts" be 0.7 and the weight of "Example Analysis" be 0.3, then the edge...<A1.1→A1.2> Dependence = 0.7 × 0.3 = 0.21. When performing the "calculation" action, the product of the label weights is used as the dependency value. The dependency threshold is set to 0.15 (obtained from a statistical approximate average; the average dependency of all label combinations in the course is set to 0.15, and this value is used as the screening criterion). If the edge dependency is lower than the threshold, the edge is removed. For example... = 0.12 then discard.<A1.1→A1.3> Edges exceeding a threshold are retained, ultimately forming a set of directed dependent paths. During this process, a "filtering" action removes all node-edge combinations with a level difference greater than 2 from the path; for example, the level difference between A1.1 and A1.4 is 3, so these are deleted.<A1.1→A1.4> Only path edges with a level difference ≤ 2 are retained to form a path set.

[0069] The number set generation submodule extracts the node numbers and sorting positions in the dependency path based on the directed dependency path set, adjusts the numbering order according to the order relationship of the nodes in the path, and combines the node position index, connection direction and path sequence to obtain the number identifier and establish the knowledge point number set.

[0070] The number set generation submodule expands upon the aforementioned path set, extracting the node numbers and their sequential positions within the paths. In specific implementations, it sequentially reads dependent paths such as...<A1.1→A1.2→A2.1> Extract the node number sequence {A1.1, A1.2, A2.1}, corresponding to the sequential positions {1,2,3}. Then, perform an "adjustment" operation on the path node positions. If a node appears in multiple paths, its position in the shortest path is used for unified numbering adjustment. For example, A2.1 in path... It is located in the 3rd position in the path. If a node is located in the second position, its numbering is uniformly adjusted to the second position. During the "judgment" action, all paths containing that node are compared, and the shortest path is selected to determine its position. Then, based on the path direction, the starting node is assigned the number K001, and subsequent nodes are numbered K002, K003, and so on. Finally, the node position index, connection direction, and path sequence are combined to generate the number identifier. For example, the path...<A1.1→A1.2→A2.1> The generated number set is K001, K002, K003, path<A2.1→A2.3> During the generation of K003 and K004, the frequency of each node in all paths needs to be counted during the "calculation" process. If the frequency of a node is greater than the average number of nodes in the path (assuming the average path length is 3), it is marked as a core node. For example, A2.1 appears 5 times and has 4 paths, so the frequency is 1.25 > 3 / 4 = 0.75, and it is marked as a core node. The node is assigned the prefix "C" first, i.e., C003, which provides the foundation for subsequent knowledge graph association.

[0071] Please see Figure 2 The learning state recognition module includes:

[0072] The behavior extraction submodule calls the knowledge point number set to obtain the answer time, score and number of modifications in each learner's answer record for each question. It performs data item standardization transformation on each type of data, classifies the standardization results in combination with the knowledge point number, and generates answer behavior association values.

[0073] Using the knowledge point ID set as an index, the system retrieves learner's answer records for each question. In practice, it first reads the corresponding knowledge point ID, answer time, score, and number of modifications for each answer record. Then, it standardizes these three data items item by item using the Z-score method, calculating the full sample mean and standard deviation for each data item. For example, taking answer time as an example, the original data is... The sample mean is calculated as follows:

[0074] The standard deviation is: The standardized value is calculated accordingly: For the set of scores and number of modifications Perform the same calculations to obtain the mean values. , Standard deviation , After calculating the standardized values ​​for each item, and combining them with the knowledge point numbers... Forming behavioral triples Each knowledge point is categorized and merged to form the final answer behavior correlation value. If a knowledge point corresponds to multiple records, the average value of each indicator is calculated according to the following formula: ,in Ultimately, standardized response behavior values ​​are generated, categorized by number.

[0075] Table 2. Raw and Standardized Data on Response Behavior

[0076]

[0077] As shown in Table 2, all data items in the answer records have been standardized and transformed, and then organized into a standardized set of behavioral vectors according to the knowledge point number, forming the basic data for subsequent state analysis.

[0078] The state matching submodule calls the answer behavior association value, compares and analyzes each data item according to the set mastery weight benchmark value and stability coefficient benchmark value, analyzes the mastery matching status of the behavior corresponding to each knowledge point, records the node state offset direction and offset magnitude, and generates a state offset coefficient set.

[0079] Based on the correlation values ​​of answering behaviors as input, the processing first requires reading three standardized behavioral values ​​corresponding to each knowledge point number: answering time, score, and number of modifications. Then, preset mastery weight benchmark values ​​and stability coefficient benchmark values ​​are invoked, and the numerical differences for each data point are calculated. Let the mastery weight benchmark value be standardized score ≥ 0.5, and the stability coefficient benchmark value be standardized modification number ≤ 0.3. Combined with the K001 behavioral vector [-0.63, 0.63, ...], [-0.39], whose score standardization 0.63 > 0.5, is judged as mastery; whose number of modifications standardization -0.39 < 0.3, is judged as stable; whose time standardization -0.63 is judged as high answering efficiency. For this matching state, a "judgment" action is performed. The rules are: if the score standardization ≥ 0.5 and the number of modifications standardization ≤ 0.3, the state match is positive mastery; if the score standardization < 0.5 and the number of modifications > 0.3, it is not mastery; if only one of these conditions is met, it is partial mastery. The state offset direction is recorded according to the direction of the current behavior value relative to the baseline value, for example... The score deviates from the baseline value in a positive direction, with an amplitude of 0.63-0.5=0.13, recorded as a positive offset. The "calculation" operation is (behavior value - baseline value), and the absolute value of the offset amplitude is taken, resulting in 0.13. Finally, the offset direction and offset amplitude under each knowledge point number are recorded and output as a set of state offset coefficients. The state of each node consists of 3 indicators, representing the mastery level offset (score), stability level offset (number of modifications), and efficiency offset (time consumption). The values ​​are all obtained through difference calculation and do not involve model calls, ensuring that the processing action is a repeatable and verifiable numerical derivation process.

[0080] The mastery classification submodule extracts the mastery weight value, stability coefficient and offset amplitude value corresponding to each node based on the state offset coefficient set. By extracting the fluctuation rate of answering time, score deviation, accuracy rate and progress rate, and using support vector machine, the learner's mastery status on each knowledge point is identified, including mastery, partial mastery and no mastery. A classification structure table is established to obtain the learning state node set.

[0081] The received state offset coefficient set is the basic data. During processing, the mastery weight value, stability coefficient, and offset amplitude value corresponding to each node are extracted. When performing a "judgment" operation on these values, a standard threshold needs to be set. A mastery weight ≥ 0.5 is judged as mastery, ∈ [0.2, 0.5) as partial mastery, and < 0.2 as no mastery. A stability coefficient ≤ 0.3 is judged as stable, otherwise as unstable. An offset amplitude ≥ 0.5 is recorded as high volatility. Combining the score offset of node K003, which is -1.05 - 0.5 = -1.55, the mastery weight is much lower than the benchmark, so it is judged as no mastery. The number of modification offset is 0.39 > 0.3, so it is judged as unstable. The state of this node is marked as no mastery - unstable type. Then, the answer time volatility corresponding to each node is extracted. Let the same knowledge The maximum and minimum time taken to answer the questions were 350s and 250s respectively, so the volatility was (350-250) / 250=0.4. The score deviation was calculated as (actual score - full score) / full score. Taking a score of 6.5 as an example, the deviation was (10-6.5) / 10=0.35. The accuracy rate was calculated as the number of correct answers / total number of answers. The improvement rate was calculated as the last score - the first score / the first score. If the first score was 6 and the last score was 8, the improvement rate was (8-6) / 6=0.33. These values ​​were used as auxiliary references for judgment. When constructing the classification structure table, the above parameters were combined to classify and label each node. The result output was a set of learning status nodes, marking each node as mastered, partially mastered, or not mastered.

[0082] Please see Figure 2 The test task generation module includes:

[0083] The master node filtering submodule calls the learning status node set, identifies nodes with the label of master status, extracts the corresponding node number and status label value, and generates a master node number set;

[0084] The learning status node set is used as the data source. During execution, all node records marked as "mastered" are first read. For each record, a "judgment" operation is performed to extract whether the status field value is "mastered". If it is "mastered", its node number information is read. For example, if the status field in the node number set is [mastered, not mastered, partially mastered, mastered], the filtering result is item 1 and item 4, corresponding to node numbers K001 and K003 respectively. Then, a new number set is constructed, which only contains node numbers with the status field "mastered". In this process, it is necessary to ensure that the same node number is recorded only once. A "filtering" operation is performed on duplicate items, and redundant numbers are removed by the set deduplication logic. The remaining node numbers are assigned a sequence label index value as a sorting reference field. When performing the "extraction" action, the status field and node field are read by field index. At the same time, double field validation is performed on the read fields to ensure that the number field is not empty and the status field accurately matches the value "mastered". Finally, the filtered node number list is organized into a number set and output as the mastered node number set.

[0085] The question ability extraction submodule calls the mastery node number set, extracts the test question information corresponding to the node, extracts the cognitive ability code and standard time data of the test question, calculates the median time value of the same type of question within the node range, statistically analyzes the time fluctuation range and ability code coverage level of the question, and obtains the cognitive feature value of the test question.

[0086] During execution, the process uses the node number set as input. First, it reads the test question records associated with each node number. For example, node K001 corresponds to questions Q101 and Q102, and node K003 corresponds to questions Q104 and Q105. It then reads the question number, cognitive ability encoding field, and standard time consumption field for each question to construct the original question feature dataset. When performing the "calculation" operation, it extracts the median time consumption data for questions under the same node number. For example, if the time consumption data for node K001 is [120, 150] seconds, the median time consumption is (120+150) / 2 = 135 seconds; if the time consumption data for node K003 is [200, 160] seconds, the median time consumption is (200+160) / 2 = 135 seconds. For 180 seconds, the number of covered questions for each cognitive ability code was statistically analyzed. When performing the "statistics" action, the cognitive code was used as the grouping basis to count the number of questions appearing under each code. For example, code A1 appeared in Q101 and Q103 with a frequency of 2, and code B2 appeared only in Q105 with a frequency of 1. After the statistics were completed, each code was sorted by frequency as a reference for the coverage level. At the same time, a "comparison" operation was performed to identify high-coverage and low-coverage codes. When dividing the interval, codes with a frequency ≥ 2 were defined as high-coverage codes, and codes with a frequency = 1 were defined as low-coverage codes. Correspondingly, A1 is a high-coverage code and B2 is a low-coverage code. Finally, the question number, corresponding node number, cognitive code, median time consumption, and coverage level were organized into a set of question cognitive feature values. The results are shown in the table below.

[0087] Table 3. Test Item Information and Cognitive Ability Coverage Table

[0088]

[0089] As shown in Table 3, the test questions under the mastery node have been categorized by node number and their cognitive ability codes and standard time consumption data have been extracted, providing a direct reference for subsequent cognitive feature collection and processing.

[0090] The priority queue construction submodule calculates the median time consumption and state weight difference for each question based on the cognitive feature value of the test questions, calculates the task sorting priority based on the state deviation and cognitive encoding complexity, filters test questions, and establishes a priority queue for push tasks.

[0091] The specific formula for calculating task priority is as follows:

[0092] ;

[0093] Calculate task sorting priority;

[0094] in, Representing the Prioritization of test questions Representing the Median time taken for each test question Representing the The state weights of the test questions This represents the average of the state weights of all test items. Representing the Test questions and the The difference in cognitive ability encoding between knowledge points The average value representing the cognitive ability codes for all test items. Represents the total number of knowledge points. Represents the currently calculated number of The identifier of each knowledge point This represents the number of each independent test question in the system;

[0095] formula:

[0096] ;

[0097] Detailed explanation of the formula and its calculation derivation:

[0098] This formula is used to calculate the priority of each test question. Test questions with higher priority values ​​will be prioritized in the push notification task. The formula provides a numerical value that measures the priority of test questions by comprehensively considering the state weight of each test question, the time spent on each test question, and the cognitive ability matching degree of each knowledge point.

[0099] Parameter meanings and settings:

[0100] For the first The state weight of the test question is set to 0.6;

[0101] The average state weights for all test questions are set to 0.5.

[0102] For the first The median time taken for each test question is set to 300 seconds.

[0103] For the first Test questions and the Cognitive ability coding difference between knowledge points, cognitive ability difference The calculation method is as follows: The dimensionless measure of the cognitive ability coding difference is because the cognitive ability coding... It is a standardized numerical value, usually within the interval [0,1], dimensionless, and does not involve physical quantities; therefore, its difference is also dimensionless. Taking the absolute value and the square root operation do not change its dimensionless nature. Let the first... The cognitive ability difference between the test question and the second knowledge point is 0.3;

[0104] The average cognitive ability code for all test items is set to 0.4.

[0105] Total number of knowledge points. Set to 10;

[0106] Substitute the parameters into the formula to calculate:

[0107] ;

[0108] ;

[0109] ;

[0110] ;

[0111] , ;

[0112] ;

[0113] ;

[0114] Interpretation of results:

[0115] Calculated priority This indicates the priority of the test question. A lower priority value means the question is relatively easy, the learner has a good grasp of it, it takes relatively little time, and it matches their cognitive ability coding well. Therefore, this question has a lower priority, and the system may prioritize recommending other test questions that are more suitable for the learner's current learning state.

[0116] Please see Figure 2 The path scheduling adjustment module includes:

[0117] The path construction submodule calls the push task priority queue, extracts the knowledge point numbers included in the task, sorts the number set in order, establishes the path connection relationship between knowledge points, configures the path identification number of the connection relationship, extracts the node sorting weight value in combination with the task order, and generates the learning path structure number set.

[0118] The task priority queue is used as the input source. During execution, the set of knowledge point numbers corresponding to each task is first extracted. Task T01 contains ["K001", "K002", "K004"], task T02 contains ["K003", "K005"], and task T03 contains ["K002", "K006"]. A sequential sorting operation is performed on each set of numbers, sorting the sets according to the task's set order weight value. The sorting criterion is from largest to smallest. For example, in task T01, the sorting weight is [0.9, ...]. [0.7, 0.5], the corresponding knowledge point numbers are arranged in order as K001→K002→K004, and then the "establish" action is executed to establish directed connection pairs between adjacent knowledge point numbers, generating a set of path connectivity relationships. For example, T01 constructs the path as K001→K002→K004, T02 as K003→K005, and T03 as K002→K006. After the connectivity relationship is established, the "configure" action is executed to assign a path identifier number to each path, for example, numbered as P001, P002, and P003 according to the task order, and the sequence position and weight of the nodes in each path are recorded. The node sorting weight is the value set in the task, which is recorded as K001-0.9, K002-0.7, K004-0.5, etc. The final structure is a set of path structure numbers composed of path number, node number, node sorting position and sorting weight, and the output is a set of learning path structure numbers.

[0119] The dependency scoring submodule calls the learning path structure number set, identifies the connection edge information of adjacent nodes in the path chain, calculates the dependency weight value between each pair of nodes according to the sorting weight, and obtains the node dependency strength score.

[0120] The specific formula for identifying the connection edge information of adjacent nodes in a path chain is as follows:

[0121] ;

[0122] Calculate the eigenvalues ​​of the complexity of connecting edges;

[0123] in, Representative node With nodes The complexity eigenvalues ​​of the edges connecting them. Represents the time slice in the path structure internal nodes With nodes Information strength value of the connecting edges This is the structural perturbation adjustment factor. Represents a node With nodes In time slice The difference in direction between inner connected edges. This is the direction weight correction factor. This is the minimum disturbance offset. This represents the total number of time slices counted in the path chain. Represents a node With nodes The edge distribution density value between them, Represents a node The mean of the edge distribution density values ​​between it and all its connected nodes. The index variable is the time slice number to be summed. This is the index number of the starting node in the path chain. This is the index number of the target node in the path chain;

[0124] formula:

[0125] ;

[0126] Detailed explanation of the formula and its calculation derivation:

[0127] Formulas are used to calculate nodes With nodes The complexity feature value of the connecting edges between adjacent nodes is used to evaluate the information complexity of the connecting edges between adjacent nodes in the path chain;

[0128] Parameter meanings and settings:

[0129] For the first Nodes in a time slice With nodes The information strength value of the connecting edges is set to 3.5;

[0130] The structural disturbance adjustment factor is set to 0.8.

[0131] For the first Nodes in a time slice With nodes The direction difference between the connecting edges is set to 2.

[0132] This is the direction weight correction factor, set to 1.2;

[0133] The minimum disturbance offset is set to 0.5;

[0134] The total number of time slices counted in the path chain is set to 3.

[0135] For nodes With nodes The edge distribution density value is obtained by measuring the distribution of edges in the network and is set to 0.6.

[0136] For nodes The mean of the edge distribution density values ​​between it and all its connected nodes is set to 0.5;

[0137] Substitute the parameters into the formula to calculate:

[0138] First time slice: , The calculation items are: Second time slice: Same as above, the calculation term is 5.5;

[0139] Third time slice: Same as above, the calculation term is 5.5;

[0140] Sum of numerators: ;

[0141] ;

[0142] Final calculation:

[0143] ;

[0144] Result 5.32 shows the node With nodes The complexity eigenvalue of the connecting edges is 5.32, which is used to evaluate the information complexity of the connecting edges between adjacent nodes in a path chain.

[0145] The path expansion submodule selects the two preceding nodes with the highest node dependency strength scores as the expansion benchmark, extracts the corresponding path numbers and expands the path chain, establishes the expanded path branch set, and generates the path expansion range.

[0146] The path extension operation is performed based on the node dependency strength score. First, the score records of all preceding nodes are read. For example, node K004's preceding nodes are K001 and K002, with scores of 0.91 and 0.88 respectively. When performing the "select" action, the score fields are compared, and the top two values ​​are taken as the base nodes for extension. The judgment criterion is score ranking: a dependency value greater than 0.85 is a strong dependency, between 0.75 and 0.85 is a medium dependency, and below 0.75 is a weak dependency. According to the above rules, node K004's preceding nodes K001 and K002 are both strong dependency nodes. Then, the selected preceding nodes are further extended... Take the path number where it is located. For example, K001 is in path P001, and K002 appears in both P001 and P003. When performing the "extract" action, copy the path segment where it is located and concatenate the current node K004 at the end to form the extended path chain K001→K002→K004, K002→K004, etc. Then perform the "build" action to generate an extended path branch set and number the new path branches. For example, P001a is formed by extending from P001, and P003a is formed by extending from P003. The final output set of path extension intervals is the extended path set plus its path source identifier and path sequence.

[0147] Table 4 Node Dependency Strength Scoring Table

[0148]

[0149] As shown in Table 4, each node completes the basic construction of the path expansion based on the two preceding nodes with the highest scores. The scores are used to determine the dependency strength and filter the expansion node chain.

[0150] Please see Figure 2 The feedback data backtracking module includes:

[0151] The answer data extraction submodule extracts the path number and node set corresponding to the extended path based on the path extension interval, collects the answer sequence data and stability coefficient corresponding to each node, classifies the answer data by node number, generates the node answer accuracy and time consumption record, and establishes the node answer change value.

[0152] Using the path expansion interval as the input starting point, the process first reads the pairing relationship between path numbers and node numbers in the expanded path set. For example, path P001 contains nodes K001, K002, and K003, path P001a contains K004, and path P003a also contains K004. Then, it reads the response sequence data for each node, including response accuracy, average time, and stability coefficient. When performing the "collection" action, it checks the response record table corresponding to each node number one by one and extracts the accuracy value, for example, K001 is 0.85, and the time taken is 130 seconds. The stability coefficient is 0.25. The above three data items are then classified and merged according to the node number. That is, if K004 appears in multiple paths, its answer records in each path need to be recorded and saved separately, and the path number field is marked. When performing the "classification" operation, the node number is used as the primary key and the path number is used as the subkey to establish a multi-value structure. After the classification is completed, the accuracy and time consumption of each node constitute the basic data of node answer changes. Then, a comprehensive form containing the fields of node number, path number, accuracy, time consumption, and stability coefficient is established to provide data sources for subsequent trend analysis.

[0153] Table 5 Node Response Data Record Table

[0154]

[0155] As shown in Table 5, node K004 was recorded multiple times because it appeared in both expansion paths. Its accuracy and time consumption differed in different paths, which were used to calculate the change in the answer trend.

[0156] The state trend calculation submodule calls the node's response change value, performs a horizontal comparison of the node's accuracy and time consumption within the path, judges and grasps the state change trend based on the direction of change, and calculates the trend score value in combination with stability parameters to obtain the learning state change trend value.

[0157] Based on the change values ​​of node responses, the processing is grouped by path number, extracting the accuracy and average time data of each node within the same path. A "comparison" operation is performed on the data between nodes to identify the direction of change of their values ​​with the path sequence. For example, in path P001, K001 has an accuracy of 0.85, K002 has 0.70, and K003 has 0.55, decreasing sequentially, indicating a negative downward trend. If the time taken is 130, 160, and 210 seconds, it indicates an upward trend. When performing the "judgment" action, if the accuracy decreases and the time taken increases, it is marked as a weakening trend; if the accuracy increases and the time taken decreases, it is marked as a strengthening trend; otherwise, it is considered a stable trend. Then, the stability coefficient of the node is read. For example, K003 has a stability coefficient of 0.45, and the stability baseline threshold is set at 0.30. If the stability coefficient is higher than 0.30, it is marked as an unstable node, and the process is executed. The "Calculation" operation converts the trend direction into a trend score. The trend score is set according to the following rules: accuracy change rate × weight a - time change rate × weight b - instability penalty c. Let a = 1.0, b = 0.5, and c = 0.2 × (stability coefficient - 0.3). Taking K001 to K003 as an example, the accuracy decrease is 0.85 - 0.55 = 0.30, and the time increase is 210 - 130 = 80 seconds. After standardization, the time change rate is 80 / 130 ≈ 0.615. The stability penalty is 0.2 × (0.45 - 0.3) = 0.03. Therefore, the trend score is -0.30 × 1 - 0.615 × 0.5 - 0.03 = -0.6375, indicating a negative shift in the learning trend. A trend score less than 0 indicates a decline in mastery, while a score greater than 0 indicates an improvement. Finally, the trend score and trend direction indicator for each node are output.

[0158] The path marker generation submodule extracts the node numbers whose trend scores exceed the state adjustment threshold based on the learning state change trend value, matches the corresponding path number information, performs structural mapping and integration of the extracted nodes and paths, generates a path update identifier structure including node number, path number and adjustment type, and establishes a path structure update instruction set.

[0159] Taking the learning status change trend value as the input, during the execution process, perform an "extraction" action on the trend score values of all nodes. Set the state adjustment threshold to ±0.5. If the trend score is greater than 0.5 or less than -0.5, trigger the path update mechanism. Record the node number information for the nodes whose score values exceed the interval. For example, if the score of K003 is -0.6375 < -0.5, it is included in the scope of change. Then, perform a "matching" operation on the node numbers to find their path number information in the original path structure. For example, K003 is located in path P001. Finally, structurally integrate the node number, path number, and adjustment type. The adjustment type is "downward adjustment" or "enhancement adjustment" to generate a path update identification record entry, such as <K003, P001, downward adjustment>. Finally, the update records of all nodes that meet the conditions form a path structure update instruction set for the system to call when updating the path;

[0160] The state adjustment threshold calculates the outlier score through statistical features within the sliding window, and setting it to exceed the set threshold triggers the learning state adjustment of the path structure.

[0161] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from a website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that includes one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0162] It should be understood that the term "and / or" in this article merely describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0163] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0164] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply 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 on the implementation process of the embodiments of the present invention.

[0165] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0166] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0167] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0168] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0169] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0170] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0171] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An information-based teaching and testing system with adaptive learning progress, characterized in that, The system includes: The knowledge node construction module extracts the course knowledge structure to obtain knowledge point nodes, constructs directed edges and generates a set of numbers, extracts teaching tags and course outline structure, uses the tag propagation algorithm to set the direction of dependency edges, constructs a directed graph structure, generates a set of knowledge point numbers and passes it to the learning status recognition module. The knowledge node construction module includes: The structure parsing submodule extracts the course knowledge structure to obtain knowledge point nodes, calls the course teaching tags and course outline structure content, identifies the tag field corresponding to each node, classifies and encodes the nodes in combination with the course structure hierarchy, identifies the topic category and index position, and establishes node distribution hierarchy values. The directed edge construction submodule calls the node distribution hierarchy value, identifies direct reference pairs between nodes based on the category and order relationship of the nodes in the structure, establishes a set of directed connections based on the reference direction and structural hierarchy, and uses a label propagation algorithm based on the label weight allocation mechanism of the course structure hierarchy to calculate the label dependency degree for each node pair, determine the propagation direction of the edge, and generate a set of directed dependency paths. The number set generation submodule extracts the node numbers and sorting positions in the dependency path based on the directed dependency path set, adjusts the numbering order according to the order relationship of the nodes in the path, combines the node position index, connection direction and path sequence to obtain the number identifier, and establishes the knowledge point number set. The learning state recognition module calls the knowledge point number set, obtains the answering behavior, matches the mastery weight and stability coefficient, records the state weight, uses the support vector machine to classify the sequence, identifies the mastery state, generates a learning state node set and passes it to the test task generation module. The test task generation module calls the learning state node set, filters the mastery state nodes, extracts the cognitive ability code and median time of the test questions, combines the state weight and the change in answering time to filter questions, constructs a priority queue for push tasks and passes it to the path scheduling and adjustment module. The path scheduling and adjustment module calls the push task priority queue, constructs a learning path chain based on the knowledge point number, extracts the path number and sorting weight, calculates the node dependency strength score, expands the path chain to form a path branch set, generates the path expansion range and passes it to the feedback data backtracking module.

2. The information-based teaching and testing system with adaptive learning progress according to claim 1, characterized in that, The knowledge point number set includes knowledge point codes, dependency edge attributes, and node graph sequence identifiers; the learning state node set includes mastery status labels, status confidence scores, and answer performance characteristics; the push task priority queue includes test question numbers, matching scoring thresholds, and scheduling order indexes; and the path extension interval includes path number sequences, branch node groups, and extended scoring indicators.

3. The information-based teaching and testing system with adaptive learning progress according to claim 1, characterized in that, The learning state recognition module includes: The behavior extraction submodule calls the knowledge point number set to obtain the answer time, score and number of modifications in each question's answer record of the learner, performs data item standardization transformation on each type of data, classifies the standardization results in combination with the knowledge point number, and generates answer behavior association values. The state matching submodule calls the answer behavior association value, compares and analyzes each data item by item according to the set mastery weight benchmark value and stability coefficient benchmark value, analyzes the mastery matching status of the behavior corresponding to each knowledge point, records the node state offset direction and offset magnitude, and generates a state offset coefficient set. The mastery classification submodule extracts the mastery weight value, stability coefficient, and offset amplitude value corresponding to each node based on the state offset coefficient set. By extracting the fluctuation rate of answering time, score deviation, accuracy rate, and progress rate, and using support vector machine, it identifies the learner's mastery status on each knowledge point, including mastery, partial mastery, and no mastery, establishes a classification structure table, and obtains the learning state node set.

4. The information-based teaching and testing system with adaptive learning progress according to claim 1, characterized in that, The test task generation module includes: The master node filtering submodule calls the learning state node set, identifies nodes labeled as mastered, extracts the corresponding node number and state label value, and generates a master node number set. The question ability extraction submodule calls the mastery node number set, extracts the test question information corresponding to the node, extracts the cognitive ability code and standard time data of the test question, calculates the median time value of the same type of question within the node range, statistically analyzes the time fluctuation range and ability code coverage level of the question, and obtains the cognitive feature value of the test question. The priority queue construction submodule calculates the median time consumption and state weight difference for each question based on the cognitive feature values ​​of the test questions, calculates the task sorting priority based on the state deviation and cognitive encoding complexity, filters test questions, and establishes a priority queue for push tasks.

5. The information-based teaching and testing system with adaptive learning progress according to claim 4, characterized in that, The specific formula for calculating the priority of the task ranking is as follows: ； Calculate task sorting priority; in, Representing the Prioritization of test questions Representing the Median time taken for each test question Representing the The state weights of the test questions This represents the average of the state weights of all test items. Representing the Test questions and the The difference in cognitive ability encoding between knowledge points The average value representing the cognitive ability codes for all test items. Represents the total number of knowledge points. Represents the currently calculated number of The identifier of each knowledge point This represents the number of each individual test question in the system.

6. The information-based teaching and testing system with adaptive learning progress according to claim 1, characterized in that, The path scheduling adjustment module includes: The path construction submodule calls the push task priority queue, extracts the knowledge point numbers included in the task, sorts the number set in order, establishes the path connection relationship between knowledge points, configures the path identification number of the connection relationship, extracts the node sorting weight value in combination with the task order, and generates the learning path structure number set. The dependency scoring submodule calls the learning path structure number set, identifies the connection edge information of adjacent nodes in the path chain, calculates the dependency weight value between each pair of nodes according to the sorting weight, and obtains the node dependency strength score value. The path expansion submodule selects the two preceding nodes with the highest scores based on the node dependency strength score and uses them as the expansion benchmark. It then extracts the corresponding path number and expands the path chain, establishes the expanded path branch set, and generates the path expansion interval.

7. The information-based teaching and testing system with adaptive learning progress according to claim 6, characterized in that, The specific formula for identifying the connection edge information of adjacent nodes in the path chain is as follows: ； Calculate the eigenvalues ​​of the complexity of connecting edges; in, Representative node With nodes The complexity eigenvalues ​​of the edges connecting them. Represents the time slice in the path structure internal nodes With nodes Information strength value of the connecting edges This is the structural perturbation adjustment factor. Represents a node With nodes In time slice The difference in direction between inner connected edges. This is the direction weight correction factor. This is the minimum disturbance offset. This represents the total number of time slices counted in the path chain. Represents a node With nodes The edge distribution density value between them, Represents a node The mean of the edge distribution density values ​​between it and all its connected nodes. The index variable is the time slice number to be summed. This is the index number of the starting node in the path chain. This is the index number of the target node in the path chain.

8. The information-based teaching and testing system with adaptive learning progress according to claim 1, characterized in that, The system also includes: The feedback data backtracking module calls the task pre-path expansion interval to obtain the path node answer sequence and stability parameters, compares the accuracy and time consumption changes, analyzes the learning state change trend, and generates an update path structure task marking instruction. The updated path structure task marking instruction includes node update marking, trend identification type, and feedback update parameters.

9. The information-based teaching and testing system with adaptive learning progress according to claim 8, characterized in that, The feedback data backtracking module includes: The answer data extraction submodule extracts the path number and node set corresponding to the extended path based on the path extension interval, collects the answer sequence data and stability coefficient corresponding to each node, performs node number classification processing on the answer data, generates node answer accuracy and time consumption records, and establishes node answer change values. The state trend calculation submodule calls the node's response change value, performs a horizontal comparison of the node's accuracy and time consumption within the path, judges and grasps the state change trend based on the change direction, and calculates the trend score value in combination with stability parameters to obtain the learning state change trend value. The path marker generation submodule extracts the node numbers whose trend score values ​​exceed the state adjustment threshold based on the learning state change trend value, matches the corresponding path number information, performs structural mapping and integration of the extracted nodes and paths, generates a path update identifier structure including node number, path number and adjustment type, and establishes a path structure update instruction set. The state adjustment threshold is calculated by outlier score based on statistical features within a sliding window. If the score exceeds the set threshold, the learning state adjustment of the path structure is triggered.

Images