Mental Health Detection — Notebooks and Data

Short and focused collection of notebooks and datasets used for mental health detection experiments (balanced/unbalanced datasets, oversampling and undersampling experiments, and evaluation metrics).

Project Contents

• kappa_calculation.ipynb — notebook to compute inter-annotator agreement (Cohen's kappa) / evaluation scripts.
• oversampled_experiments.ipynb — experiments using oversampling techniques.
• undersampled_experiments.ipynb — experiments using undersampling techniques.
• model_results_balanced.csv — model output / metrics for balanced training.
• model_results_unbalanced.csv — model output / metrics for unbalanced training.
• train_data_balanced.csv — balanced training dataset.
• train_data_unbalanced.csv — unbalanced training dataset.
• test_data.csv — test dataset used for evaluation.
• manual_annotation_data.csv — human-annotated data used for validation.

Quick Setup

1. Clone the repository:

git clone <repo-url>
cd mental_health_detect

2. Create and activate a Python virtual environment:

python -m venv venv
source venv/Scripts/activate

3. Install dependencies:

pip install -r requirements.txt

Run the Notebooks

Start Jupyter and open the notebooks in a browser:

jupyter notebook

Open the notebook you want (for example oversampled_experiments.ipynb) and run the cells top-to-bottom to reproduce experiments and figures.

Notes

• The repository contains Jupyter notebooks and CSV data files.
• requirements.txt contains commonly used packages inferred from the notebooks. If you run into a missing package error, install the package shown in the error and re-run.
• If you prefer Docker or conda, adapt the environment creation to your preferred workflow.