CN-121707796-B - Learning strategy evaluation and guidance method based on experimental operation sequence mode mining

CN121707796BCN 121707796 BCN121707796 BCN 121707796BCN-121707796-B

Abstract

The invention discloses a learning strategy evaluation and guidance method based on experimental operation sequence pattern mining. The method comprises the steps of collecting an original operation log and reconstructing the original operation log into a normalized operation event sequence, extracting silence interval characteristics between adjacent operations, analyzing cognitive state labels corresponding to the silence intervals in combination with operation contexts to generate a semantic enhancement type operation sequence, then identifying repeated operation fragments in the sequence, calculating exploration intention characteristics to distinguish scientific exploration and random try intention, extracting strategy feature vectors of students in combination with a typical strategy mode library, further positioning key decision branch points, extracting decision micro-behavior characteristics to calculate decision hesitation indexes, evaluating strategy efficiency, planning a progressive transition path from a current strategy to a target strategy in combination with real-time cognitive load level, and generating personalized guiding suggestions. The invention can realize deep quantitative analysis and accurate guidance of the cognitive state, the operation intention and the decision psychology of the students.

Inventors

ZHAO MIN
YU XIN
LI JIQIANG
TAO KUN

Assignees

南京百伦斯智能科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260214

Claims (5)

1. The learning strategy evaluation and guidance method based on experimental operation sequence pattern mining is characterized by comprising the following steps of: collecting original operation log data generated by students on an experiment platform, and reconstructing the original operation log data into a normalized operation event sequence according to an operation time stamp; Extracting silent interval features between adjacent operations in a normalized operation event sequence, analyzing a cognitive state label corresponding to the silent interval by combining the operation context features, embedding the cognitive state label into the normalized operation event sequence, and generating a semantic enhancement type operation sequence; Identifying repeated operation fragments in the semantic enhancement type operation sequence, calculating exploration intention characteristics to distinguish operation intention types, and extracting strategy feature vectors of students by combining a typical strategy pattern library; Positioning key decision branch points in the semantic enhancement type operation sequence, extracting decision micro-behavior characteristics in a decision process to calculate decision hesitation indexes, and generating a strategy efficiency evaluation result by combining strategy characteristic vectors; Calculating a real-time cognitive load level based on the semantic enhancement type operation sequence, planning a progressive transition path from a current strategy to a target strategy by combining a strategy efficiency evaluation result, and outputting personalized strategy guiding suggestions; Extracting operation context characteristics corresponding to each silencing interval, wherein the operation context characteristics comprise complexity grade of a preamble operation, type conversion relation between a follow-up operation and the preamble operation and relative position proportion of the current silencing interval in the sequence, wherein the type conversion relation comprises similar continuation, cross-class switching and reverse rollback, and the silencing interval characteristics are composite characteristics of fusion silencing interval duration and operation context characteristics; the method comprises the steps of embedding a cognitive state label into a normalized operation event sequence to generate a semantic enhancement type operation sequence, wherein the cognitive state label comprises a cognitive state label node, a corresponding cognitive state label node, a semantic enhancement type operation sequence and a semantic enhancement type operation sequence, wherein the cognitive state label node comprises a resolved cognitive state label, a corresponding silencing interval duration and membership vectors of different cognitive states; The method comprises the steps of identifying repeated operation fragments in a semantic enhanced operation sequence, namely scanning the semantic enhanced operation sequence, and identifying continuous operation subsequences of the same type, wherein when the length of the operation subsequences of the same type exceeds a threshold value and the numerical value of operation parameters in the subsequences is changed, the operation subsequences of the same type are marked as the repeated operation fragments; The method comprises the steps of calculating exploration intention characteristics to distinguish operation intention types, wherein the exploration intention characteristics comprise multi-dimensional characteristics of each repeated operation segment to be taken as exploration intention characteristics, wherein the exploration intention characteristics comprise parameter change regularity characteristics, observation behavior characteristics, parameter space coverage characteristics and parameter space coverage characteristics, wherein the parameter change regularity characteristics are used for representing monotone trend and step length consistency of operation parameter changes in segments; The method comprises the steps of calculating monotonicity index and step length consistency index of parameter change of repeated operation fragments, calculating exploration intention index in a weighted summation mode based on monotonicity index, step length consistency index, observation behavior characteristic and parameter space coverage characteristic, judging the operation intention type of the repeated operation fragments as scientific exploration when the exploration intention index is larger than a classification threshold, otherwise judging as random trial.
2. The method of claim 1, wherein extracting the student's policy feature vector in combination with the canonical policy pattern library comprises: Matching the semantic enhanced operation sequence with the operation intention type with a plurality of typical strategy modes in a typical strategy mode library respectively; For each typical strategy mode, calculating the subsequence inclusion degree, the operation sequence consistency and the cognitive rhythm similarity of the student sequence and the typical strategy mode; Generating a matching score based on the sub-sequence inclusion degree, the operation sequence consistency and the cognitive rhythm similarity, and combining the matching scores for each typical strategy mode into a strategy feature vector.
3. The method of claim 1, wherein extracting decision micro-behavioral features in the decision process comprises: scanning a semantic enhancement type operation sequence, and positioning key decision branch points of students facing multipath selection; setting a decision window aiming at each key decision branch point, and extracting multi-dimensional decision micro-behavior characteristics in the decision window; The decision microaction features include the number of previews of alternatives before making the final selection, the number of times the operation is cancelled or rolled back after initiating, the frequency of switching between different options, and the dwell time before making the final selection.
4. A method according to claim 3, characterized in that the calculation of the decision hesitation index comprises in particular: Carrying out standardized processing on the decision micro-behavior characteristics; and calculating a decision hesitation index based on the standardized decision micro-behavior characteristics, wherein the decision hesitation index is positively contributed by the cancelling or rollback times, the switching frequency and the pause time corrected based on the cognitive state, and the decision hesitation index is negatively contributed by the previewing times.
5. The method of claim 4, wherein the corrected dwell time based on the cognitive state comprises: retrieving a cognitive state label corresponding to the pause time before final selection in the semantic enhanced operation sequence; If the cognitive state label is the target thinking, reducing the contribution weight of the pause time when calculating the decision hesitation index; if the cognitive state label is confusion stagnation, the contribution weight of the pause time length in calculating the decision hesitation index is improved.

Description

Learning strategy evaluation and guidance method based on experimental operation sequence mode mining Technical Field The invention relates to the technical field of education data mining and intelligent coaching, in particular to a learning strategy evaluation and guidance method based on experimental operation sequence mode mining. Background In the virtual simulation experiment teaching, the exploration process of students is deeply analyzed, and the intelligent teaching is realized. Through deep mining of massive operation log data generated by students on an experimental platform, cognitive paths of the students can be restored from a microscopic level, and thinking strategies of the students in links of hypothesis verification, variable control, result thinking and the like are quantitatively analyzed. The technical means can break through the limitation that the traditional teaching only pays attention to experimental results, construct an intelligent guide system (ITS) with self-adaptive capacity, and realize real-time diagnosis and intervention of the student learning process. At present, experimental teaching assessment techniques focus on rule-based outcome decisions and shallow behavioral statistical analysis. The existing scheme acquires operation logs of students from a system background, and counts the completion degree of operation steps, the accuracy of experimental conclusion and the total time length of task completion. In terms of process analysis, the prior art applies a traditional sequence pattern mining algorithm (such as a prefix projection sequence pattern mining algorithm PrefixSpan or an association rule mining algorithm Apriori) to extract high-frequency operation subsequences, or simply divides the time interval between operations into long pauses and short pauses by setting a global fixed time threshold, so as to serve as a basis for cutting operation tasks or judging whether students are active. However, the prior art primarily treats the oplog as a discrete sequence of symbols, ignoring the underlying deep cognitive semantics underlying in the gaps of operation and microscopic behavior, resulting in coarse granularity and inaccurate intent recognition of policy evaluations. Accordingly, further research and innovation are needed to solve the above-mentioned problems of the prior art. Disclosure of Invention In view of the above problems of the prior art, the application provides a learning strategy evaluation and guidance method based on experimental operation sequence pattern mining. According to one aspect of the application, a learning strategy evaluation and guidance method based on experimental operation sequence pattern mining comprises the following steps: collecting original operation log data generated by students on an experiment platform, and reconstructing the original operation log data into a normalized operation event sequence according to an operation time stamp; Extracting silent interval features between adjacent operations in a normalized operation event sequence, analyzing a cognitive state label corresponding to the silent interval by combining the operation context features, embedding the cognitive state label into the normalized operation event sequence, and generating a semantic enhancement type operation sequence; Identifying repeated operation fragments in the semantic enhancement type operation sequence, calculating exploration intention characteristics to distinguish operation intention types, and extracting strategy feature vectors of students by combining a typical strategy pattern library; Positioning key decision branch points in the semantic enhancement type operation sequence, extracting decision micro-behavior characteristics in a decision process to calculate decision hesitation indexes, and generating a strategy efficiency evaluation result by combining strategy characteristic vectors; And calculating a real-time cognitive load level based on the semantic enhancement type operation sequence, planning a progressive transition path from the current strategy to the target strategy by combining a strategy efficiency evaluation result, and outputting personalized strategy guiding suggestions. The method has the beneficial effects that deep quantitative analysis and accurate guidance on the cognitive state, the operation intention and the decision psychology of students can be realized. The related art effects will be described in detail below in connection with specific embodiments. Drawings Fig. 1 is a flowchart of a learning strategy evaluation and guidance method based on experimental operation sequence pattern mining according to an embodiment of the present application. FIG. 2 is a flowchart of identifying duplicate operation segments in a semantically enhanced operation sequence according to an embodiment of the present application. FIG. 3 is a flow chart of the method for distinguishing operation intention types by calculating explorati