Non Linear ICA and Logistic Regression for Classification of Epilepsy from EEG Signals

One of the age old neurological disorders found in human beings is epilepsy. It is witnessed by the abnormal electrical activities in the brain which causes recurring seizures, therefore the patient suffers from loss of consciousness. Due to the random nature of the seizures, the patients may not be aware of it and so it increases the risk of physical injury. Due to the disturbed brain activity, epileptic seizures are caused which implies that the brain's normal activity is suddenly interrupted and changed. Electroencephalogram (EEG) is used widely for capturing the recordings and impulses in the brain. These signals are highly useful for checking the abnormalities not just related to epilepsy but any neurological disorder. The recordings of the EEG signals are quite long in nature and so processing it is very difficult and consumes a lot of time. So in this paper, the dimensions of the EEG signals are reduced with the help of Non Linear Independent Component Analysis (ICA) and then the dimensionally mitigated values are classified through Logistic Regression Classifier. The result analysis shows that an average Performance Index of about 91.39%, an average accuracy of about 95.88 along with a less time delay of 2.085 seconds is obtained.