Depression Detection on Social Media Using Transfer Learning

This project predicts depression using Reddit comments. By leveraging machine learning, natural language processing (NLP), and embeddings, it identifies depression-related text patterns.

Features

• Dataset: eRisk dataset of Reddit comments.
• Text Preprocessing: Cleaning, lemmatization, and tokenization.
• Feature Extraction:
  • Pre-trained embeddings: Google Word2Vec, Facebook FastText.
  • Custom-trained embeddings.
• Imbalance Handling: SMOTE for synthetic oversampling.
• Model: Bidirectional LSTM with regularization and dropout.

Performance

Metric	Value
Accuracy	62.44%
Precision	58.73%
Recall	83.98%
F1-Score	69.12%
AUC-ROC	62.42%
AUC-PR	75.36%

Installation

1. Clone the repository:

   bash

   Copy code

   git clone https://github.com/username/repo-name.git

2. Navigate to the directory:

   bash

   Copy code

   cd repo-name

3. Download pre-trained models:

   • GoogleNews Word2Vec
   • Facebook FastText

Usage

1. Preprocessing:

   python

   Copy code

   python preprocess.py

2. Training:

   python

   Copy code

   python train_model.py

3. Evaluation:

   python

   Copy code

   python evaluate_model.py

Enhancements

Future work includes:

• Incorporating transformer models like BERT.
• Adding metadata-based features.
• Improving hyperparameter optimization.

Contributors

• Hossein

License

This project is licensed under the MIT License.