Epilepsy Checker

We used EEG Data to predict whether an individual has epilepsy or not. We used a publicly available dataset from UCI Machine Learning Repository. The EEG Data was noisy and the feature space was quite large (still had 178 features in the preprocessed data). Thus, we standardized the data and then used Dimensionality Reduction to reduce the feature space. Specifically, we used Principal Component Analysis (PCA) and then further analyzed the feature importances to pick the 20 most important features. After that, we used RandomizedSearchCV to find the best hyperparameters for the XGBoost model and trained the model on 70% of the dataset. The model had ~96% accuracy on the test set and 97% on the training set. The F-Score, Precision, and Recall were all >95%. After achieving that performance we exported the Model. We created a website using React for the frontend and Flask for the backend where individuals can enter the values of the 20 features in the pre-processed EEG Data that correspond to the EEG channels and we used the ML Model in the backend to predict if they have epilepsy or not.

Built With

Share this project:

Updates