A Motor Activity Analysis as a Depression Indicator: Predictive Approach Using a Hybrid CNN and LSTM Network

Depression is a prevalent mental disorder affecting behavior, cognition, and physical activity. Early detection and intervention are crucial to mitigating its severe impacts, including suicidal ideation or suicide attempts. This study proposes a novel predictive model using a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network to analyze motor activity data as an indicator of depression. Utilizing the “Depresjon” dataset, which comprises activity levels from both depressed and non-depressed individuals, the model extracts features through CNN and processes these with LSTM to identify depression patterns. To address class imbalance in the dataset, Synthetic Minority Over-sampling Technique (SMOTE) was employed, enhancing the model's predictive accuracy. The final model achieved a high accuracy rate of 97.89%, significantly outperforming the other model created. This research demonstrates the potential of combining CNN and LSTM for effective depression prediction, providing a valuable tool for early diagnosis and intervention.