Accurate detection of epileptic seizure activity from Electroencephalogram (EEG) data is crucial for the timely intervention and treatment. Studies reveal that various methodologies have been used for this purpose. However, this work presents a machine learning-based approach for classifying averaged EEG signal into ictal (seizure) and non-ictal (non-seizure) stages using a Random Forest (RF) and Support Vector Machine (SVM) classifiers. This work aims at creating a framework to automate the process of monitoring EEG signals acquired from patients suffering from epilepsy disorders. The process involves filtration of signal using a Band Pass filter for noise removal, averaging, spike detection using adaptive thresholding, clustering, and classification of EEG signal into ictal and non-ictal stages of the signal. The EEG dataset is bench-marked dataset with record varies from patient to patient, making it essential to develop an algorithm that can automatically label the signal into its corresponding stages. We propose a method that makes use of K-Means Clustering for pre-labelling the records along with an SVM/RF classifier to identify ictal and non-ictal stages. The study demonstrates the use of averaged signal and classical ML methods together serve as a computationally efficient and interpretable method for classifying ictal and non-ictal states. Results indicate both SVM (76.74% accuracy) and RF (83.72% accuracy) effectively determine the seizure and non-seizure states with RF showing marginal superiority compared to SVM.