CN-121981554-A - Academic risk diagnosis method based on multi-time scale learning evolution analysis

Abstract

The invention discloses a academic risk diagnosis method and system based on multi-time scale learning evolution analysis. The learning method comprises the steps of obtaining learning behavior and learning result time sequence data formed by students in a teaching system, carrying out long-term learning trend analysis under a first time scale to determine trend types, carrying out recent learning dynamic analysis under a second time scale to determine dynamic types, taking the long-term trend types and the recent dynamic types as combined input, and outputting corresponding academic risk state types based on preset mapping rules. The system comprises a data management module, a trend analysis module, a dynamic analysis module, a diagnosis rule engine module and an intervention linkage module. The invention can accurately identify the evolution process of academic risk, has good interpretability and applicability, supports layering analysis and automatic intervention triggering, and is suitable for various online teaching and teaching management scenes.

Inventors

• Guo Zehan

Assignees

• 牡丹江师范学院

Dates

Publication Date

20260505

Application Date

20260205

Claims (10)

1. 1. The academic risk diagnosis method based on multi-time scale learning evolution analysis is characterized by comprising the following steps of: acquiring time sequence data of learning behaviors of a target student and related learning results; Based on the time series data, carrying out long-term learning trend analysis under a first time scale, and determining a long-term learning trend type; Based on the time series data, performing a recent learning dynamic analysis under a second time scale, and determining a recent learning dynamic type by comparing a recent learning change characteristic with the long-term learning trend type; Taking the long-term learning trend type and the recent learning dynamic type as combined input, and determining the academic risk state type uniquely corresponding to the combination through a pre-established diagnosis mapping rule so as to complete academic risk diagnosis; Wherein the second time scale is smaller than the first time scale; the first time scale is a time scale spanning multiple teaching periods or periods, and the second time scale is a time scale of the last one or more teaching periods.
2. 2. The method according to claim 1, wherein the recent learning dynamics analysis uses the aforementioned long-term learning trend type as a constraint condition, and the recent learning dynamics type is determined by comparing a direction consistency and a change amplitude relationship between a recent learning change characteristic and a long-term learning trend.
3. 3. The method of claim 1 or 2, wherein the long-term learning trend type comprises at least one of a continuously ascending type, a continuously descending type, a wave-like variation type, and a relatively stable type.
4. 4. The method of claim 1 or 2, wherein the recent learning dynamics type comprises at least one of a collaborative change, a deviating change, a turning change, and no significant change.
5. 5. The method of claim 4, wherein the deviation change is further classified as either a strong deviation or a weak deviation based on a proportional relationship between a recent change magnitude and a long-term change magnitude.
6. 6. The method of claim 1, wherein the diagnostic mapping rules are deterministic combination mapping rule tables, different long-term learning trend types corresponding to unique academic risk status types in combination with recent learning dynamic types.
7. 7. The method of claim 6, wherein the academic risk status type comprises at least one of a steady development type, a fluctuating risk type, a progressive deterioration type, a sudden risk type, a risk mitigation type, and a recovery development type.
8. 8. The method according to claim 1, wherein the academic risk diagnosis method is completed based on only time-series data formed by learning behavior and learning results, independent of student psychological assessment data or artificial subjective assessment results.
9. 9. The method of claim 1, further comprising the step of triggering a learning intervention strategy recommendation, an early warning prompt or a monitoring frequency adjustment corresponding thereto based on the determined academic risk status type.
10. 10. A academic risk diagnosis system, comprising: One or more processors; a memory storing computer program instructions; The computer program instructions, when executed by the one or more processors, cause the system to perform the academic risk diagnosis method based on multi-time scale learning evolution analysis of any one of claims 1 to 10.

Description

Academic risk diagnosis method based on multi-time scale learning evolution analysis Technical Field The invention relates to the field of education data mining and intelligent learning analysis, in particular to a academic risk diagnosis method based on multi-time scale learning evolution analysis. Background Under the current informatization background of education, an online learning platform and a teaching management system are widely applied to higher education and middle and primary school teaching scenes, and behavior data and achievement results of students in the learning process are systematically recorded and stored. The data are used for carrying out academic risk identification and early warning, and become an important means for improving teaching management efficiency and guaranteeing student academic development. Traditional academic risk recognition methods rely on static performance evaluation or behavior indexes in a single time window, such as single examination scores, learning duration of a week, job submission conditions and the like. Such methods typically model academic risk identification as a classification or regression prediction problem, using machine learning or statistical models to directly output "risk" labels. However, this type of approach suffers from the following significant drawbacks: On one hand, the method ignores the time evolution characteristic of the student learning process, and cannot comprehensively describe the student state from a dynamic angle. The student's academic risk is often not caused by a single abnormal behavior, but rather is formed by the evolution trend of behavior over a period of time in combination with a sudden change of state over a short period of time. Therefore, the risk determination at a single point in time or time window may be distorted, and it is difficult to reflect the actual evolution process of the student status. On the other hand, the existing method lacks modeling of the association relation between the long-term learning trend and the recent learning dynamics in the analysis model structure, and is difficult to reveal the staged characteristics and the potential mechanism of the student risk formation. For example, a student may have a long-term downward trend but recently developed behavioral rebound, and may suddenly develop significant abnormalities in the last few weeks, with the overall performance being stable. The conventional model often gives the same risk judgment for the two cases, and cannot distinguish between a 'continuous worsening type' and a 'burst abnormality type' risk state. In addition, the interpretation capability of part of the model on the risk type is insufficient, the output result of the model lacks clear meaning, and teaching management personnel cannot easily understand the judgment basis, so that targeted intervention measures are difficult to take based on the result. Furthermore, some models require the introduction of student psychological assessment questionnaires, interview recordings, or subjective assessment data by the teacher, which not only increases the cost of data acquisition, but also limits the versatility of the system. In summary, the existing academic risk identification technology has the following core problems: 1. the evolution modeling capability of the learning process under multiple time scales is lacking, and the change of the academic state caused by different risk mechanisms is difficult to accurately diagnose; 2. The lack of a definite trend and dynamic joint analysis structure leads the risk judgment result to be output by a model, and has poor interpretation and operability; 3. The data dependence is high, and the stable deployment in a large-scale teaching platform is difficult. Therefore, a technical solution based on a multi-time scale analysis framework, having a hierarchical diagnosis structure and outputting a academic risk status type with definite semantics is needed, so as to improve accuracy, interpretation and applicability of risk identification, and provide scientific support for personalized learning intervention and accurate teaching decision. Disclosure of Invention In view of the above, the invention aims to provide a learning industry risk diagnosis method based on multi-time scale learning evolution analysis to realize accurate modeling of student learning industry state change process and output risk types with definite semantics, thereby improving the interpretability, stability and practical application value of diagnosis results. In order to achieve the above purpose, the present invention provides the following technical solutions: In one embodiment of the invention, a academic risk diagnosis method based on multi-time scale learning evolution analysis is provided, comprising the steps of acquiring time series data of learning behaviors of a target student and related learning results, carrying out long-term learning trend analysis under a