Epilepsy is considered as one of the most common neurological disorder that affects people of all ages, races and ethnic backgrounds.Epileptic seizures are characterized by an unpredictable occurrence pattern and transient dysfunctions of the central nervous system, due to excessive and synchronous abnormal neuronal activity in the cortex.To diagnose epilepsy, EEG signal interpretation is considered as the most prominent testing tools because it is painless,low cost, and has efficient temporal resolution of long-term monitoring.However, to develop an efficient method, the main challenges are designing an appropriate feature extraction method and selecting the most prominent features because the quality of feature set plays an important role on the classification accuracy.In this thesis, we addressed the issue of feature selection for the machine leaming using a correlation based approach.Our main hypothesis is that good feature sets consist of features that are highly correlated with the class; however, they are not correlated with each other.
　　We have developed a method for feature selection named as Improved Correlation Based Feature Selection (ICFS).Our proposed method was implemented in the publicly available EEG data set.First we tested this method in binary classification, where we showed that whether the signal was epileptic or not.In the second phase we evaluated our proposed method in multiclass domain, where we classified a signal as ictal, interictal and healthy.For the multiclass problem we named the feature selection method as Extended Correlation Based Feature Selection (ECFS).
　　After the feature selection process, the reduced feature space was applied to five different classification algorithms as Random Forest, Support Vector Machine, Naive Bayes, K-Nearest Neighbor and Logistic Model trees.Our experimental results showed that the proposed method would be a promising candidate method for the classification of epileptic seizure.