A whole-process learning situation tracking and evaluation system for interdisciplinary theme learning

Abstract

This invention discloses a full-process learning tracking and evaluation system for interdisciplinary thematic learning, belonging to the field of educational informatization and intelligent learning evaluation. It includes a multimodal full-process data perception module, an interdisciplinary three-dimensional coupled learning modeling module, a time-series learning evolution intelligent engine module, a value-added evaluation and diagnosis module, and a multi-role scenario-based interaction module. This invention covers all stages of learning, employing a seamless acquisition architecture combining edge and cloud computing to acquire structured, semi-structured, and unstructured multimodal data; constructing a unified learning ontology that three-dimensionally couples subject knowledge, interdisciplinary literacy, and the learning process; generating a dynamic learning map through a multi-source heterogeneous fusion engine; using a time-series graph neural network to drive learning evolution tracking; and incorporating algorithms for ability transfer path identification and cognitive breakpoint localization; performing value-added evaluation based on an individual's initial baseline; constructing a traceable evidence chain; and outputting hierarchical diagnosis and graded adaptive intervention schemes.

Description

Technical Field

[0001] This invention relates to the field of educational informatization and intelligent learning assessment technology, and in particular to a full-process learning tracking and assessment system for interdisciplinary thematic learning. Background Technology

[0002] Interdisciplinary thematic learning is a core learning model under the background of the new curriculum reform. It is driven by real problem situations, integrates knowledge, thinking and methods from two or more disciplines, and guides students to solve complex problems through inquiry, collaboration, practice and reflection. Its core goal is to cultivate students' higher-order core competencies such as systematic thinking, innovative practice and collaborative communication.

[0003] Current mainstream learning assessment systems are generally built around single-subject knowledge systems, and have the following shortcomings when adapting to the assessment needs of interdisciplinary thematic learning: First, the assessment has a single dimension and is knowledge-based, which can only quantify the mastery of knowledge points in a single discipline. It cannot track and evaluate higher-order goals such as interdisciplinary core competencies and ability transfer in an observable and quantifiable way, thus being out of touch with the training objectives of interdisciplinary learning. Secondly, there is a serious lack of process data. It relies on structured summative data such as tests, assignments, and final results, but lacks the collection of multimodal data on behavior, cognition, and emotion throughout the entire process of inquiry, collaboration, practice, and reflection. It is unable to capture the ability development nodes and cognitive obstacles in the learning process. Third, the learning profile is static, and can only output phased static assessment results. It cannot depict the dynamic evolution of learning over time, nor can it identify the ability transfer path in interdisciplinary scenarios. It is difficult to locate cognitive breakpoints at the intersection of disciplines, and the diagnostic depth is insufficient. Fourth, the assessment logic is rigid and highly subjective. It often uses a uniform scale to conduct cross-group benchmarking, ignoring the differences in individual student baselines. At the same time, the assessment of interdisciplinary learning outcomes relies heavily on teachers' subjective judgment and lacks a traceable process evidence chain to support it, resulting in insufficient objectivity and fairness in the assessment conclusions. Fifth, multi-source data suffers from the silo effect, with data from learning platforms, collaboration tools, experimental systems, and classroom data collection devices being fragmented and unable to form a complete cross-disciplinary learning profile, leading to one-sided evaluation conclusions.

[0004] Therefore, a comprehensive learning tracking and assessment system for interdisciplinary thematic learning is needed to meet the needs of full-process, multi-dimensional, and developmental assessment of interdisciplinary thematic learning. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a full-process learning tracking and evaluation system for interdisciplinary thematic learning, solving the technical problems of existing learning evaluation systems, such as single evaluation dimensions, missing process data, static learning profiles, strong evaluation subjectivity, prominent data silos, and inability to adapt to the literacy-oriented evaluation needs of interdisciplinary thematic learning.

[0006] Technical Solution: To solve the above-mentioned technical problems, according to one aspect of the present invention, more specifically, a system for tracking and evaluating the entire learning process of interdisciplinary thematic learning, including a multimodal full-process data perception module, an interdisciplinary three-dimensional coupled learning modeling module, a time-series learning evolution intelligent engine module, a value-added assessment and diagnosis module, and a multi-role scenario-based interaction module. Details are as follows: The multimodal full-process data perception module covers the entire learning stage of interdisciplinary thematic learning, including pre-learning preparation, inquiry practice, collaborative discussion, output of results, and reflection and iteration. It collects structured, semi-structured, and unstructured multimodal learning data, employing a non-intrusive, seamless data collection architecture that combines edge computing and cloud computing to achieve standardized data reporting. The collected multimodal learning data includes structured assessment scores and task progress data, semi-structured collaborative interaction and operation log data, and unstructured audio and video, experimental operations, learning outcomes, and retrieval behavior data. Specifically, the seamless data collection architecture involves local preprocessing and non-intrusive data collection at the classroom end via intelligent recording edge devices, unified format data reporting via SDK-embedded learning, collaboration, and experimental tools at the online end, and multi-source data aggregation and standardized cleaning at the cloud end.

[0007] The interdisciplinary three-dimensional coupled learning learning modeling module is used to construct a unified learning learning ontology that couples three dimensions: subject knowledge, interdisciplinary literacy, and learning process. Through a multi-source heterogeneous data fusion engine, multimodal learning data is mapped to corresponding nodes of the learning learning ontology, generating a three-dimensional coupled dynamic learning learning map with temporal attributes. Specifically, the three dimensions of the interdisciplinary three-dimensional coupled learning learning ontology are: subject knowledge dimension, which constructs knowledge point entities based on multidisciplinary curriculum standards and labels them with subject attributes and interdisciplinary cross-relationships; interdisciplinary literacy dimension, which includes six primary literacy indicators (systematic thinking, critical innovation, problem-solving, communication and collaboration, information literacy, and practical application) and corresponding quantifiable secondary indicators; and learning process dimension, which includes four process-oriented state indicators (learning engagement, cognitive load, learning strategies, and reflective ability). The entities and indicators of the three dimensions are coupled and mapped through association edges, allowing the same learning behavior data to simultaneously correspond to knowledge nodes, literacy indicators, and process indicators. The multi-source heterogeneous data fusion engine adopts entity alignment technology and confidence-weighted fusion mechanism to automatically map multimodal data from different sources and in different formats to the corresponding nodes of the three-dimensional learning ontology, and dynamically assigns fusion weights according to the confidence of the data source to correct misjudgments of learning caused by the bias of a single data source.

[0008] The time-series learning learning evolution intelligent engine module is used to take the three-dimensional coupled dynamic learning learning map as the input of the dynamic graph structure, and update the attribute weights of each node in the map and the strength of the association edges between nodes in real time through the time-series graph neural network model, and output the time-series evolution curves of learning learning in each dimension. The time-series learning learning evolution intelligent engine module has built-in interdisciplinary ability transfer path identification algorithm and interdisciplinary cognitive breakpoint identification algorithm, which are used to quantify the degree of ability transfer between different subject tasks and output the transfer path, locate the location and cause of cognitive obstacles at the interdisciplinary knowledge nodes.

[0009] The time-series graph neural network model uses knowledge nodes, literacy indicators, and process indicators in the three-dimensional coupled dynamic learning learning graph as dynamic graph nodes, and the correlation between indicators as graph edges. It dynamically updates the node attribute values ​​and the connection strength of the edges by combining the learning data input in time series, generates continuous time-series evolution curves of learning learning in each dimension, and identifies the causal relationship between different learning learning indicators.

[0010] The interdisciplinary ability transfer path identification algorithm calculates the degree of interdisciplinary ability transfer of students from subject task A to subject task B using the following formula. :

[0011] In the formula: The Pearson correlation coefficient is used to measure the score sequence of students in subject task A and subject task B, respectively, to reflect the covariance consistency of ability performance. The cosine similarity between two subject tasks in terms of cognitive operation, problem context, and knowledge domain features is calculated from the pre-trained task feature vectors. The normalized value of the shortest path distance between knowledge nodes of subject A and subject B in the three-dimensional learning ontology is used to represent the structural knowledge barriers between subjects. The adjustment parameters for the strength of the impact of knowledge barriers are set based on experience in scenarios involving interdisciplinary crossover. The interdisciplinary ability transfer path identification algorithm quantifies the degree of student ability transfer between different subject tasks based on the interdisciplinary ability transfer metric, visualizes the ability transfer path map, and identifies key promoting and hindering nodes for ability transfer.

[0012] The interdisciplinary cognitive breakpoint identification algorithm combines interdisciplinary knowledge nodes in the three-dimensional learning map, matches behavioral data such as the student's dwell time at the corresponding node, the error rate of answering questions, the number of times resources were repeatedly consulted, and the semantic features of the question content, automatically locates the specific position of the cognitive breakpoint, and classifies and identifies the cause type of the cognitive breakpoint.

[0013] The value-added assessment and diagnosis module is used to establish a three-dimensional initial learning baseline for individual students before the thematic learning begins. Based on this individual initial baseline, it calculates the learning value-added magnitude in each dimension in stages, constructs a traceable evidence chain with a one-to-one correspondence between assessment conclusions and process data, outputs stratified diagnostic results, and generates a graded adaptive intervention plan. Specifically, the three-dimensional initial learning baseline is a student's initial learning profile covering knowledge, literacy, and process dimensions, generated through pre-assessment and baseline surveys before the thematic learning begins. The learning value-added magnitude is calculated based on the student's individual initial baseline, with longitudinal growth values ​​for each dimension calculated according to project milestone stages, without relying on cross-group benchmarking results as the core assessment. The traceable evidence chain links each assessment conclusion to corresponding process data segments supporting that conclusion, enabling full-link traceability and verification of the assessment conclusions. The tiered diagnostic results include knowledge-level diagnosis, competency-level diagnosis, and strategy-level diagnosis, which are used to locate weaknesses in single-subject knowledge and blind spots in cross-disciplinary knowledge, identify shortcomings and growth bottlenecks in cross-disciplinary core competencies, and analyze process problems in learning methods and collaborative strategies, respectively. The graded adaptive intervention program is divided into three levels: individual level, group level, and class level, which respectively match personalized learning resources, collaborative role optimization programs, and teaching adjustment suggestions.

[0014] The multi-role scenario-based interaction module is used to output corresponding dimensions of learning data, assessment results, and intervention suggestions to different user roles, realizing the scenario-based implementation of assessment results. The multi-role scenario-based interaction module includes teacher-side units, student-side units, teaching and research units, and management units. The teacher-side unit displays a whole-class learning dashboard, dynamic group collaboration profiles, individual growth reports, and anomaly warning lists. The student-side unit displays personal growth profiles, ability radar charts, visualization of ability transfer paths, and reflection guidance scaffolds. The teaching and research unit conducts course effectiveness attribution analysis and course optimization direction analysis. The management unit displays an overview of interdisciplinary learning and teaching quality monitoring reports.

[0015] The beneficial effects of this invention's interdisciplinary thematic learning full-process learning tracking and evaluation system are as follows: (1) This invention realizes the standardized fusion mapping of multi-source heterogeneous learning data and the systematic learning modeling of cross-disciplinary dimensions by constructing a multimodal, non-intrusive acquisition system covering the entire stage of interdisciplinary learning and a unified learning ontology that couples knowledge, literacy and process in three dimensions, thus breaking down data silos among multiple tools; By using a time-series neural network-driven dynamic learning evolution tracking and a quantifiable interdisciplinary ability transfer computing mechanism, we have achieved continuous dynamic characterization of learning and in-depth diagnosis of interdisciplinary ability growth, breaking through the limitations of traditional static profiling in depicting complex learning processes. By establishing an evaluation mechanism that combines baseline benchmarking, value-added calculation, and evidence chain tracing, we have achieved individual-oriented developmental assessment and verifiable objective evaluation, thus solving the pain points of strong subjectivity and insufficient fairness in interdisciplinary assessment.

[0016] This invention effectively breaks through the technical bottlenecks of existing single-subject, summative, and static assessment systems. It can accurately adapt to the competency-oriented assessment needs of interdisciplinary thematic learning, significantly improve the comprehensiveness, objectivity, and diagnostic depth of learning assessment, and provide reliable technical support for the optimization of interdisciplinary teaching and the cultivation of students' higher-order competencies. Attached Figure Description

[0017] The present invention will now be described in further detail with reference to the accompanying drawings and specific implementation methods.

[0018] Figure 1 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0019] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0020] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1: Application scenario of interdisciplinary thematic learning on "Carbon Neutrality and Urban Development" in junior high school This example is designed for a four-week interdisciplinary thematic learning program on "Carbon Neutrality and Urban Development" for junior high school students. It integrates four subjects: geography, chemistry, mathematics, and moral education and the rule of law. The core driving problem is "designing local urban carbon emission reduction pathways." The specific operation process of this system is as follows: Before the thematic learning begins, the value-added assessment and diagnosis module generates a three-dimensional initial learning baseline for each student, covering knowledge, literacy, and process, through pre-knowledge assessment, literacy scale survey, and learning status questionnaire. It records the students' initial levels in carbon emission-related knowledge points, problem-solving abilities, and learning engagement, which serve as the benchmark for subsequent value-added assessments.

[0022] During the thematic learning process, the multimodal, full-process data perception module covers all stages, including pre-construction preparation, inquiry and practice, collaborative discussion, output of results, and reflection and iteration. It employs a seamless edge + cloud acquisition architecture to acquire multimodal data: in the classroom, intelligent recording edge devices collect data on classroom voice interactions, student attention levels, and facial expressions, pre-processing the data locally before uploading it to the cloud; online, an SDK is embedded in the learning platform, collaborative discussion tools, and virtual simulation experiment system to collect resource access trajectories, group discussion texts, mind map nodes, experimental operation logs, answer data, and search behavior data; simultaneously, unstructured data such as student progress reports and result demonstration videos are also collected. All data is aggregated, standardized, cleaned, and formatted in the cloud.

[0023] The interdisciplinary 3D coupled learning learning modeling module constructs subject knowledge dimension nodes based on the curriculum standards of four subjects, and marks interdisciplinary cross-links such as "carbon emission accounting" and "climate system change". It configures 6 primary literacy indicators such as system thinking and problem solving and corresponding secondary indicators. Combined with 4 process indicators such as learning engagement, it constructs a 3D coupled learning learning ontology. The multi-source heterogeneous data fusion engine maps various types of collected data to corresponding nodes of the learning learning ontology through entity alignment technology, and uses a confidence weighting mechanism to correct the deviation of different data sources, generating a 3D coupled dynamic learning learning map with time-series attributes.

[0024] The time-series learning literacy evolution intelligent engine module takes a three-dimensional coupled dynamic learning literacy map as input. Through a time-series graph neural network model, it dynamically updates the attributes of each node and the strength of related edges over time, generating continuous evolution curves of students' learning literacy across various dimensions. This identifies the causal relationship that "improved information literacy leads to improved problem-solving efficiency." The interdisciplinary ability transfer path identification algorithm calculates the transferability of students' abilities from the "chemical carbon emission accounting task" to the "geographical climate impact analysis task": it calculates the Pearson correlation coefficient based on the student's performance scores in five trials for each of the two tasks. cosine similarity of features between the two types of tasks The normalized distance of the shortest path to the knowledge nodes corresponding to the two types of tasks in the learning ontology. Adjust parameters Take 0.8 and substitute it into the formula to calculate the mobility. Based on this, the system quantifies the degree of students' variable control and data analysis abilities transfer from chemistry to geography, identifying knowledge barriers as the main obstacles to this transfer. An interdisciplinary cognitive breakpoint identification algorithm, combined with student behavioral data, pinpoints cognitive breakpoints at the interdisciplinary intersection of "carbon emission accounting - climate impact prediction," caused by insufficient interdisciplinary knowledge connections. Simultaneously, the system uses an abnormal learning situation early warning algorithm to identify abnormal states of cognitive overload in some students, triggering warnings.

[0025] At weekly milestones, the value-added assessment and diagnosis module uses each student's initial baseline as a benchmark to calculate the learning value-added in each dimension. Each assessment conclusion is linked to the corresponding process data to form a traceable chain of evidence. It outputs three-level hierarchical diagnostic results at the knowledge, literacy, and strategy levels, and generates tiered intervention plans: at the individual level, interdisciplinary learning resources are pushed to students with cognitive gaps; at the group level, role optimization suggestions are pushed to groups with insufficient collaborative participation; and at the class level, teaching adjustment suggestions for common cognitive problems are pushed to teachers.

[0026] The multi-role scenario-based interaction module provides corresponding functions for different roles: Teachers can view the whole class learning dashboard, dynamic profiles of group collaboration, and anomaly warning lists; students can view their personal growth profiles, ability radar charts, and visualization results of ability transfer paths; the teaching and research side can conduct attribution analysis of the effectiveness of thematic courses and identify directions for course optimization; and the management side can view the overall situation of interdisciplinary learning in the grade.

[0027] Example 2: Application Scenario of Interdisciplinary Project-Based Learning in High School "Smart Community Water Conservation Solution Design" This example is designed for a 6-week interdisciplinary project-based learning program for high school students on "Designing Smart Community Water-Saving Solutions." It integrates four subjects: physics, general technology, mathematics, and information technology. The core task is to "design feasible smart community water-saving renovation solutions." The specific operation process of this system is as follows: Before the project started, the value-added assessment and diagnosis module established a three-dimensional initial learning baseline for each student through pre-existing technical assessments, practical literacy surveys, and learning strategy questionnaires, recording the initial levels of fluid mechanics knowledge, practical application ability, engineering thinking, and other dimensions.

[0028] During project implementation, the multimodal full-process data perception module covers the entire learning stage, adopting an edge + cloud architecture to collect multimodal data: in the classroom and laboratory, intelligent acquisition devices are used to acquire videos of physical experiment operations, equipment operation data, and classroom discussion audio; online, project management tools, programming platforms, and design tools are embedded through an SDK to collect task progress, code operation logs, solution design process data, group collaboration records, and information retrieval behavior; at the same time, unstructured results data such as field survey records, final design schemes, and prototype test data are collected, and data aggregation and standardization are completed in the cloud.

[0029] The interdisciplinary 3D coupled learning learning modeling module constructs knowledge dimension nodes corresponding to four disciplines, and marks cross-related elements such as "fluid mechanics calculation" and "water-saving algorithm modeling". It configures 6 primary literacy items and corresponding secondary indicators, and combines them with 4 process indicators to form a 3D coupled learning learning ontology. The multi-source heterogeneous data fusion engine maps various practical and technical data to ontology nodes to generate a 3D coupled dynamic learning learning map with time-series attributes.

[0030] The time-series learning evolution intelligent engine module dynamically updates the learning map through a time-series graph neural network model, outputting a learning evolution curve and identifying the synergistic growth relationship between practical application literacy and problem-solving ability. The interdisciplinary ability transfer path identification algorithm calculates the transferability of students' abilities from the "physical water pipe flow measurement experiment task" to the "information technology water-saving algorithm modeling task": Pearson correlation coefficients are calculated based on the students' performance scores for each of the two tasks, taken four times each. Cosine similarity of features between the two types of tasks Normalized distance of the shortest path to knowledge nodes Adjust parameters Take 1.0 and substitute it into the formula to calculate the mobility. This study quantifies the degree to which students' data measurement and error control abilities can be transferred from physics experiments to information technology modeling, identifying differences in disciplinary thinking as obstacles to this transfer. An interdisciplinary cognitive breakpoint identification algorithm pinpoints cognitive breakpoints in students at the intersection of "measured water data - algorithm model parameters," attributing this to a lack of data transformation methods.

[0031] At the mid-term and final milestones of the project, the value-added assessment and diagnosis module calculates the value-added magnitude of each dimension based on the initial baseline, constructs a full-chain evidence, and outputs three-level diagnostic results and graded intervention plans: at the individual level, it pushes specialized learning resources on data transformation methods to students; at the group level, it pushes role optimization plans for groups with unreasonable division of labor in the project; and at the class level, it provides supplementary suggestions for teachers to provide practical scaffolding.

[0032] The multi-role scenario-based interactive module provides teachers with dynamic monitoring of project progress and group collaboration, students with personal practical ability growth files and reflection guidance, teaching and research staff with attribution analysis of project-based learning effectiveness, and management staff with quality statistics of practical interdisciplinary courses.

[0033] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A system for tracking and evaluating learning progress throughout the entire process of interdisciplinary thematic learning, characterized in that, It includes a multimodal full-process data perception module, an interdisciplinary three-dimensional coupled learning situation modeling module, a time-series learning situation evolution intelligent engine module, a value-added assessment and diagnosis module, and a multi-role scenario-based interaction module; The multimodal full-process data perception module is used to cover the entire learning stage of interdisciplinary thematic learning, including pre-learning preparation, exploration practice, collaborative discussion, output of results, and reflection and iteration. It collects structured, semi-structured and unstructured multimodal learning data and completes standardized data reporting using a non-intrusive and seamless data collection architecture that combines edge computing and cloud computing. The interdisciplinary three-dimensional coupled learning situation modeling module is used to construct a unified learning situation ontology with three dimensions of subject knowledge, interdisciplinary literacy, and learning process. Through a multi-source heterogeneous data fusion engine, multimodal learning data is mapped to the corresponding nodes of the learning situation ontology to generate a three-dimensional coupled dynamic learning situation map with time-series attributes. The time-series learning ability evolution intelligent engine module is used to take the three-dimensional coupled dynamic learning ability map as the input of the dynamic graph structure, and update the attribute weights of each node in the map and the strength of the association edges between nodes in real time through the time-series graph neural network model, and output the time-series evolution curves of learning ability in each dimension; the time-series learning ability evolution intelligent engine module has built-in interdisciplinary ability transfer path recognition algorithm and interdisciplinary cognitive breakpoint recognition algorithm, which are used to quantify the degree of ability transfer between different subject tasks and output the transfer path, locate the location and cause of cognitive obstacles at the interdisciplinary knowledge nodes; The value-added assessment and diagnosis module is used to establish a three-dimensional initial learning baseline for individual students before the start of thematic learning, calculate the learning value-added of each dimension in stages based on the individual initial baseline, construct a traceable evidence chain that corresponds one-to-one with the assessment conclusions and process data, output hierarchical diagnostic results and generate a graded adaptive intervention plan. The multi-role scenario-based interaction module is used to output learning data, assessment results and intervention suggestions of corresponding dimensions to different user roles, so as to realize the scenario-based implementation of assessment results.

2. The system for tracking and evaluating the learning progress of interdisciplinary thematic learning as described in claim 1, characterized in that, The multimodal learning data collected by the multimodal full-process data perception module includes structured assessment scores and task progress data, semi-structured collaborative interaction and operation log data, and unstructured audio and video, experimental operations, learning outcomes and retrieval behavior data. The aforementioned seamless data acquisition architecture specifically involves local preprocessing and non-intrusive data acquisition at the classroom end via intelligent recording edge devices, unified format data reporting at the online end via SDK-embedded learning, collaboration, and experimental tools, and standardized data aggregation and cleaning at the cloud end.

3. The whole-process learning tracking and evaluation system for interdisciplinary thematic learning according to claim 1, characterized in that, The three dimensions of the interdisciplinary three-dimensional coupled learning ontology are: subject knowledge dimension, which constructs knowledge point entities based on multidisciplinary curriculum standards and marks subject attributes and interdisciplinary cross-relationships; Interdisciplinary literacy includes six primary literacy indicators: systems thinking, critical innovation, problem solving, communication and collaboration, information literacy, and practical application, as well as corresponding quantifiable secondary indicators. The learning process dimension includes four process-oriented state indicators: learning engagement, cognitive load, learning strategies, and reflective ability. The entities and indicators in the three dimensions are coupled and mapped through the associated edges, so that the same learning behavior data can be mapped to knowledge nodes, literacy indicators and process indicators at the same time.

4. The whole-process learning tracking and evaluation system for interdisciplinary thematic learning according to claim 1, characterized in that, The multi-source heterogeneous data fusion engine adopts entity alignment technology and confidence-weighted fusion mechanism to automatically map multimodal data from different sources and in different formats to the corresponding nodes of the three-dimensional learning ontology, and dynamically assigns fusion weights according to the confidence of the data source to correct misjudgments of learning caused by the bias of a single data source.

5. The whole-process learning tracking and evaluation system for interdisciplinary thematic learning according to claim 1, characterized in that, The time-series graph neural network model uses knowledge nodes, literacy indicators, and process indicators in the three-dimensional coupled dynamic learning learning graph as dynamic graph nodes, and the correlation between each indicator as graph edges. It dynamically updates the node attribute values ​​and edge connection strengths by combining the learning data input in time series, generates continuous time-series evolution curves of learning learning in each dimension, and identifies the causal relationships between different learning learning indicators.

6. The whole-process learning tracking and evaluation system for interdisciplinary thematic learning according to claim 1, characterized in that, The interdisciplinary ability transfer path identification algorithm calculates the degree of interdisciplinary ability transfer of students from subject task A to subject task B using the following formula. : In the formula: The Pearson correlation coefficient is used to measure the score sequence of students in subject task A and subject task B, respectively, to reflect the covariance consistency of ability performance. The cosine similarity between two subject tasks in terms of cognitive operation, problem context, and knowledge domain features is calculated from the pre-trained task feature vectors. The normalized value of the shortest path distance between knowledge nodes of subject A and subject B in the three-dimensional learning ontology is used to represent the structural knowledge barriers between subjects. The adjustment parameters for the strength of the impact of knowledge barriers are set based on experience in scenarios involving interdisciplinary crossover. The interdisciplinary ability transfer path identification algorithm quantifies the degree of student ability transfer between different subject tasks based on the interdisciplinary ability transfer metric, visualizes the ability transfer path map, and identifies key promoting and hindering nodes for ability transfer.

7. The whole-process learning tracking and evaluation system for interdisciplinary thematic learning according to claim 1, characterized in that, The interdisciplinary cognitive breakpoint identification algorithm combines interdisciplinary knowledge nodes in the three-dimensional learning map, matches behavioral data such as the student's dwell time at the corresponding node, answer error rate, number of times resources were repeatedly consulted, and semantic features of the question content, automatically locates the specific position of the cognitive breakpoint, and classifies and identifies the cause type of the cognitive breakpoint.

8. The whole-process learning tracking and evaluation system for interdisciplinary thematic learning according to claim 1, characterized in that, The three-dimensional initial learning baseline is a student's initial learning profile covering three dimensions: knowledge, literacy, and process, generated through pre-assessment and baseline survey before the start of thematic learning. The learning value-added is based on the student's initial baseline and the longitudinal growth value of each dimension is calculated according to the project milestone stage, without taking the results of cross-group benchmarking as the core of the evaluation. The traceable evidence chain consists of each evaluation conclusion being associated with a corresponding process data segment that supports that conclusion, enabling full-chain backtracking verification of the evaluation conclusion.

9. A system for tracking and evaluating the learning progress of interdisciplinary thematic learning as described in claim 1, characterized in that, The tiered diagnostic results include knowledge-level diagnosis, competency-level diagnosis, and strategy-level diagnosis, which are used to locate weaknesses in single-discipline knowledge and blind spots in cross-disciplinary knowledge, identify shortcomings and growth bottlenecks in cross-disciplinary core competencies, and analyze process problems in learning methods and collaboration strategies, respectively. The tiered adaptive intervention program is divided into three levels: individual, group, and class, which respectively match personalized learning resources, collaborative role optimization plans, and teaching adjustment suggestions.

10. A system for tracking and evaluating the learning progress of interdisciplinary thematic learning as described in claim 1, characterized in that, The multi-role scenario-based interaction module includes a teacher's end unit, a student's end unit, a teaching and research end unit, and a management end unit; The teacher-side unit is used to display a dashboard of the whole class's learning progress, dynamic portraits of group collaboration, individual growth reports, and an anomaly warning list; The student-side unit is used to display personal growth profiles, competency radar charts, visualizations of competency transfer paths, and reflection guidance scaffolds. The teaching and research unit is used to conduct attribution analysis of course effectiveness and analysis of course optimization directions; the management unit is used to display an overview of interdisciplinary learning and teaching quality monitoring reports.

Images