Machine Learning-Based Classification for Effective Recognition and Examination of Epileptic Seizures

Recurring seizures are a hallmark of the neurological disease epilepsy, which severely diminishes the standard of living for those afflicted. Roughly half a million individuals throughout the globe suffer with epilepsy, a long-term condition marked by repeated convulsions. Timely treatment and better results may be achieved with the early diagnosis of seizures through the analysis of medical pictures. When trained on EEG and MRI data, deep learning models have shown encouraging results in the detection and classification of epilepsy. Continuous efforts are needed to improve the accuracy and reliability of these models in order to achieve excellent performance in epilepsy detection. Nonetheless, the study shows a number of ways to make deep learning models better at identifying epilepsy. The input data is first cleaned and prepared using data pre-processing techniques, such as noise reduction, data normalization, and managing missing values. The training data is further enhanced in terms of variety and generalizability by means of data augmentation techniques including random translations, scaling, and rotations. Secondly, alternative model topologies are investigated in an effort to enhance the model’s epilepsy detection capabilities. Also, to make the deep learning model work better, hyperparameter adjustment is used. When compared to other machine learning classifiers, the suggested naïve bayes (NB) approach detects epileptic seizures with the greatest accuracy $98.4 \%$. Some more approaches include stochastic gradient descent (SGD), logistic regression (LR), and K-nearest neighbors (KNN). When it comes to identifying seizures, NB is by far the best of them.