Research on the Trend of Students' Ideological Dynamics in Civic Education Based on Time Series Prediction Modeling

Abstract

This paper takes students' daily behavioral habits as the entry point to explore the changes of students' ideological dynamics in ideological education through behavioral analysis. The behavioral trajectory of students' school life is divided into five modules: course learning, school life, social interaction, employment and ideology and politics, and this is used to form a behavioral dataset of the research sample. After fusing the data from multiple sources to obtain the spatial and temporal fine-grained daily activity trajectories of the samples, the mathematical formulas for students' daily activities, daily behaviors and daily behavioral habits on campus are designed on the basis of which, as well as the calculation method for the intensity of the habits, the modeling method for students' daily behavioral habits is proposed. The extracted behavioral time feature sequences are input into the multivariate time series classifier, and the MLSTM-FCN model is determined to be the best model. Since its FCN branch has the problem of small sensory field and only supports single-item operation when extracting the sequence features, the Transformer encoder is used to replace the FCN unit, and the model sensory field is enlarged by introducing the form of multi-attention mechanism, so as to establish the time-series-based student abnormal behavior detection model. The model is calculated to get the intensity of the Civic Learning Resources Attention Behavior of Normal Behavior Students as 41.23% in the analysis of the change of students' ideological dynamics in Y college. It summarizes students' ideological cognition and behavioral changes through time series features, reflects students' growth patterns and effective education pathways, and points out the direction for updating and improving the model of civic education in colleges and universities.