The ongoing development of various lightweight and portable EEG signal acquisition devices provides the opportunity to implement home-based epilepsy monitoring. However, it is essential to apply a highly effective method to handle the limited computational power of such devices. In this paper, we propose a cost-effective method to classify epileptic seizure using stratified sampling technique. Additionally, to reduce the required computational power, this paper proposes a novel correlation and threshold-based feature selection algorithm. For evaluating the performance of our proposed method, five different classification algorithms are applied to classify the epileptic seizure from the reduced feature set. In our experiment, the random forest classifier shows the highest accuracy compared to other classifiers.