Published Paper Details:

Bipolar disorder is a chronic and debilitating psychiatric illness characterized by recurrent episodes of mania and depression, significantly affecting an individual's emotional stability, cognitive functioning, and quality of life. Accurate diagnosis remains challenging due to symptom overlap with other mood disorders and reliance on subjective clinical assessments. This study presents a clinically driven deep learning framework for the automated prediction of bipolar disorder using a structured, symptom-based dataset reflecting common psychiatric indicators. Three deep learning models--Artificial Neural Network (ANN), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM)--were designed and evaluated to analyze behavioral and clinical symptom patterns. Standard preprocessing techniques, including normalization and train-test splitting, were applied to ensure reliable model performance. The predictive capability of each model was assessed using classification accuracy and confusion matrix analysis. Results demonstrate that deep learning approaches effectively model complex clinical relationships within psychiatric data. The ANN model achieved the highest predictive accuracy, indicating its suitability for tabular clinical datasets. This work emphasizes the potential of deep learning as a supportive clinical decision-making tool for early bipolar disorder screening and improved mental health assessment.