Epileptic Seizure Recognition using Machine Learning Algorithms

The project entitled “Epileptic Seizure Recognition Using Machine Learning Algorithms” is used to identify the epileptic seizure at the early stage. Epileptic seizure is a significant health concern affecting millions worldwide, necessitate precise and timely identification for effective medical intervention and management. The present system for epileptic seizure recognition often relies on traditional methods, including expert analysis of electroencephalogram (EEG) data and manual review. While these methods have proven effective, they are labor-intensive, prone to subjectivity and may not be suitable for real-time monitoring. This gap led to the proposed machine learning-based approach which offer potential seizure recognition, allowing for more timely and accurate intervention. This study propose a novel approach for epileptic seizure recognition by combining dimensionality reduction techniques, namely Principal Component Analysis (PCA) and t-Distributed Stochastic Neighbor Embedding (t-SNE), with the K-Means clustering algorithm, then integrate this unsupervised clustering approach with four popular machine learning algorithms: Support Vector Machine classifier, Logistic Regression, Naive Bayes classifier and Random Forest classifier to enhance the performance of seizure recognition. Each cluster is labeled as either seizure or non-seizure based on the majority class within it. The method involves the development and implementation of these algorithms within the context of epileptic seizure. The primary objective is to explore the feasibility and performance of these classifiers, aiming to enhance the efficiency of seizure recognition system. The work emphasizes the potential of machine learning in automating the identification of epileptic seizures, ultimately contributing to improved patient care and medical diagnosis. The improved accuracy of $95 \%$ is achieved by Random Forest model.