Epilepsy is one of the brain based disease affects human due to over electricity power passed in the brain. It is also a neurological disorder problem comes after stroke or brain fever, brain attack or while less blood flow in the brain. Recurrent attack creates epilepsy mainly. Number of death increased nowadays due to epilepsy seizures. In order to control this it is essential to identify and detect the epilepsy seizures from Electroencephalography (EEG) recorded from the patients. Certain number of symptoms led a doctor to suspect epilepsy seizures. One of the most important tools to diagnose neurological problems and epilepsy seizures is by analysing the EEG. Therefore EEG is considered as an indispensable tool for diagnosing epilepsy in medical industry. The occurrence of the epilepsy is irregular and it is difficult to predicting the seizures automatically. Hence, in this paper it is motivated to design and develop a novel approach for detecting and classifying the epilepsy-seizures from EEG signals automatically. The existing research work presents an approximate entropy method and improved approximate entropy method. But the obtained accuracy is very less. In order to improve the accuracy of the classification, this paper aimed to provide a novel approach which comprises of a sequence of signal processing steps such as: filtering, amplifying, decomposing, feature extraction and selection and classification. The experiment is carried out in MATLAB software and the results investigated and evaluated by comparing with the existing system results. The proposed approach obtained 12% of accuracy is improved than the existing approaches.