An electroencephalogram (EEG) is a recording of the electrical activity of the brain. EEG signals are mainly used for the detection and visualization of many neurological disorders. It is also used for the diagnosis of epilepsy. Epilepsy is a chronic disease and 1% of the world's population is affected. Epilepsy results in seizures associated with loss of consciousness, including long-term loss of consciousness and affecting patient quality of life. It is possible to determine the detection of epilepsy from EEG data. However, there are some difficulty to detect epilepsy because of the length and complexity of EEG. There are many studies for the classification and detection of epilepsy. Most of these studies are realized with artificial neural networks and machine learning. In this study, epilepsy detection was performed by using machine learning over EEG data with the features obtained by time-frequency domain analysis during seizure. As a result of the study, Logistic Regression classifier gave the best training results for 23 different EEG channels and tested. In the test results; 87.5% accuracy, 75.0% sensitivity and 100% specificity values were obtained.