Electroencephalogram (EEG) is the most commonly used signals in the detection of Epilepsy. EEG video monitoring can help the neuro-physician to detect and diagnose the epilepsy however; the workload of neuro-physician can be reduced by automatic epilepsy detection. Aim of proposed work is to introduce the combination of statistical, spatial and two novel spectral features Bispectrum Magnitude Variance and Bispectrum Magnitude Average in pair for optimizing the process of epileptic detection. We got promising result 99.41% as the classification accuracy, through various features on multichannel EEG, than traditional algorithms. Selecting pair-wise multi-domain features have improved the accuracy and efficiency of a classifier from 96.47 to 99.4%. The outflow information of the multi-domain and multichannel features was considered as the input vectors to the SVM for discriminating the epileptic signal from the normal EEG. Features selected are very useful for region wise localization of epilepsy.