| Summary: | This paper focuses on the design of an educational early warning mechanism based on the fusion of ideological education and multi-featured data so as to manage the educational situation of students in colleges and universities efficiently and accurately. In this paper, the wavelet transform, discrete Fourier transform, and lag sequence analysis algorithms are used to effectively extract temporal features of students’ behaviors. PageRank and Hit’s algorithms are employed to extract features related to student concept maps. The emotional tendencies recognition interface provided by Tencent Cloud was used to obtain the emotional features of students’ speeches. Following this, a multi-feature fusion was performed to depict the students’ learning. A Hive-based data warehouse is used to integrate heterogeneous data from multiple sources. Finally, the education early warning model based on multi-feature data fusion is introduced, and the operation mechanism of early warning mechanism for ideological and political education in colleges and universities is established. To verify the effect of this paper’s model against other algorithms, this paper’s model achieves the optimal performance in the F1 score in negative samples, which is 0.91, followed by the TPA-LSTM algorithm, which is 0.88. Before the optimization of the early warning mechanism, the average per capita absenteeism of the students was 1.32 sessions, and the rate of disciplinary actions was 0.0291. At the end of the academic year, the average per capita absence rate decreases to 1.24 sessions, and the disciplinary action rate decreases to 0.0245.
|
|---|