COMPARING ML MODELS TO REDUCE MISDIAGNOSIS IN PSYCHIATRIC DISORDERS USING EEG DATA

This project aimed to reduce the misdiagnosis of psychiatric disorders by leveraging machine learning algorithms on EEG data. Misdiagnosis in psychiatry can lead to ineffective treatment plans and prolonged patient suffering. By utilizing EEG data, which captures electrical activity in the brain, this project sought to identify patterns that could differentiate between various psychiatric disorders more accurately.

Algorithms Used:

• LightGBM: A gradient boosting framework that uses tree-based learning algorithms, known for its speed and efficiency.
• CatBoost: A high-performance gradient boosting library that handles categorical features seamlessly.
• Logistic Regression: A statistical model that in this context serves as a baseline, providing a probabilistic approach to binary classification problems.

Key Outcomes:

• The project compared the performance of LightGBM, CatBoost, and Logistic Regression in classifying psychiatric disorders based on EEG data.
• LightGBM and CatBoost demonstrated superior accuracy and robustness compared to the baseline Logistic Regression model.
• The use of EEG data proved effective in highlighting distinctive brain activity patterns associated with different psychiatric disorders, thereby improving diagnostic accuracy.