Mental Health Predictor: A Machine Learning & LLM-Based System for Mental Health Condition Detection and Explanation

📌 Project Overview:

Mental health is a crucial aspect of well-being, yet early detection and awareness remain challenging. This project presents an AI-driven mental health condition predictor that:

• Classifies user input into categories such as Depression, Anxiety, Stress, Bipolar, Personality Disorder, Suicidal, or Normal.
• Generates natural language explanations for the detected condition using a fine-tuned Mistral model.
• Provides coping mechanisms and potential next steps based on the predicted mental health condition.
• Features a Streamlit-based UI for seamless user interaction.

🚀 Key Features:

• Text-Based Classification: Predicts mental health conditions from user input.
• LLM-Generated Explanations: Uses a fine-tuned Mistral model for natural language explanations.
• Coping Mechanisms & Suggestions: Provides recommendations to manage symptoms.
• User-Friendly Interface: Interactive Streamlit UI for easy testing.
• Open-Source & Expandable: Built using free, open-source LLMs and models for transparency and scalability.

🔧 Tech Stack & Tools:

• Natural Language Processing (NLP): Fine-tuned Mistral (via Ollama) & transformer models (T5)
• Machine Learning Frameworks: PyTorch, Hugging Face Transformers
• UI Implementation: Streamlit
• Model Training & Experimentation: Google Colab
• Version Control & Deployment: GitHub

📖 How to Use:

1️⃣ Clone the Repository

git clone https://github.com/ArshdeepKaur04/Mental-Health-Predictor.git

cd mental-health-predictor

2️⃣ Install Dependencies

After cloning the repo, users can install all dependencies with:

pip install -r requirements.txt

3️⃣ Run the Streamlit App

streamlit run ui/app.py

Enter text describing your symptoms, and the model will predict the condition, generate an explanation, and suggest coping mechanisms.

🏆 Results & Findings:

The model has been fine-tuned using labeled mental health datasets, and results show high accuracy in detecting conditions. The fine-tuned Mistral model effectively generates contextual explanations and relevant coping strategies.

🛠 Future Improvements:

• Enhance dataset with more diverse and real-world inputs.
• Optimize response personalization using retrieval-augmented generation (RAG).
• Expand UI to include interactive visualization of mental health trends.
• Integrate with mental health APIs for additional resources.

🤝 Contributing:

Contributions are welcome! Feel free to fork the repository, raise issues, or submit pull requests.

📝 License:

This project is open-source under the MIT License.

🔗 Large Files (Model, Video & Project Documentation):

Download trained models and the demo video here: https://drive.google.com/drive/folders/1Qv1-XGQkH_Nyibz2v_A3fVqL7ZDRbnN6?usp=sharing