Abstract
Adapting to new academic and social settings is a major problem for first-year students in university. This problem has become increasingly more pressing due to the increasing complexity of societal, spiritual, and economic shifts. When it comes to helping students adjust to college life and thrive academically and personally, no one does a better job than universities and colleges. Data preparation, PCA, and model training make up the three stages of this study's methodology. The feature selection process used min-max normalisation, and principal component analysis was used to lower the dimensionality of the input and output indicators. The suggested architecture combines CNNs for hierarchical feature extraction and LSTMs for collecting long-term dependencies in temporal data; both networks work in tandem to achieve this goal. At 203.64% MAE, 241.22% RMSE, and 1.78% MAPE, the CNN-LSTM model fared better than the alternatives. This research proves that the model is useful for assessing college students' actions. Finally, the study highlights how important it is to use advanced machine learning techniques to help students adjust and how teachers may play a bigger role in creating meaningful learning experiences.