Project Title: Predicting Mental Health and Mood Swings Using Deep Learning
Impact Statement: Applied AI to mental health research, focusing on early detection of mood swings to enable personalized interventions.
Problem: Mood swings can serve as early indicators of mental health challenges such as anxiety and depression. Traditional assessments often fail to capture subtle, complex patterns influenced by demographic, lifestyle, and emotional factors.
Approach:
Data: Used the Kaggle Mental Health Dataset, including features like age, gender, employment status, sleep quality, social activity, and self-reported stress, anxiety, and depression.
Preprocessing: Cleaned data by handling missing values, encoding categorical variables, and normalizing continuous features.
Model: Developed a deep neural network with ReLU activation layers to capture non-linear relationships between inputs and mood swings.
Training: Optimized the network with iterative training, hyperparameter tuning, and validation to ensure robust predictions.
Results:
Achieved 83% accuracy in predicting mood swings.
Emotional factors, particularly stress and anxiety, were identified as stronger predictors than demographic data.
The model demonstrated balanced performance across different mood categories, validated through F1-score, precision, and recall metrics.
Future Directions:
Integrate longitudinal datasets to track mood changes over time.
Incorporate wearable and mobile sensor data for real-time monitoring.
Collaborate with clinical practitioners to validate predictions and support applied interventions.
Impact:
This project demonstrates how AI and deep learning can provide a scalable, data-driven approach to mental health research, supporting early diagnosis and personalized care strategies for improved societal well-being.
This PDF paper was published in multiple science journals, including JSHS and Curiex.