Deep Learning Hybrid Models for Detecting Stress Biomarkers Among EFL Teachers in Higher Education

EFL instructors in universities, whose students have exceptionally high expectations, are particularly vulnerable to the negative effects of occupational stress on their emotional and physical well-being. Nearly half of all first-year educators leave the industry due to the high stress levels, which puts teaching among the top 30 most stressful professions. To improve the well-being of both educators and students, it is vital to identify biomarkers of physiological and psychological stress in this population. In order to classify pertinent data and analyse occupational stress, this study utilised a number of sophisticated DL models, including CNN-SVM, VAE-RNN, CNN-RNN, CNNRNN, and a hybrid CRNN-LSTM. By incorporating a memory cell to store learning at each classification interval, the CRNNLSTM model outperformed the others, with a classification accuracy of 95.53%. The analysis uncovered 42 important topics, with the utmost emphasis on Learning Needs, Higher Education, and Social Impact. The presence of stress biomarkers was significantly linked with these themes.