© 2018, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 1354 The Design of Algorithm for Analysis of EEG Datasets for Prediction of Epilepsy Aprajita Sambyal 1 , Dr. Bikram Pal Kaur 2 , Amanpreet Kaur 3 1 Student, Dept.of IT., CEC,Mohali(PB),India 2 Professor, Dept. of IT., CEC,Mohali(PB),India 3 IKJPTU, Jalandhar(PB),India -------------------------------------------------------------------------***---------------------------------------------------------------------- Abstract- The study and analysis of sleep Electroencephalogram (EEG) data has long been done to understand the psychology and brain function among patients and individuals. The data collected from these studies have helped researchers in arriving at various well documented conclusions. In this work, the evaluation of EEG data on different perspectives shall be done so that the brain waves and EEG hypnograms can be evaluated with predictive mining. A unique model of Improved ANN shall be devised and trained from the dataset fetched. We have used epilepsy based data to work with the predictive model with Random Forest on EEG Data. Key Words: EEG(Electroencephalograph) ,Epilepsy, Machine Learning, Physionet, Polyman, Python, Random Forest, SVM(Support Vector Machine) 1. INTRODUCTION The new age has provided mankind with a lot of technology to ease the life of an individual but this so called period of DzSmart-productsdz has also made man more dependent on technology and has led to a rise in stress and anxiety. This stress and anxiety manifests itself as various brain disorders in humans. Our aim is to study the sleep EEG signals captured from patients’ brain and study these to perform predictive analysis on them. We design ANN and train it with sample datasets captured from Physionet. The proposed technique is an improved hybrid approach for prediction called Random Forest Approach. 2. STEPS OF PROPOSED WORK  Extraction of Live EEG Data from Bioinformatics Research Portals including Physionet.  Transformation of EEG Data to ASCII Format using execution in Polyman and Python.  Training of Soft Computing Model from ASCII data so that predictions can be done on testing data.  Development and Execution of Python based Implementation on Pre-Processed EEG-ASCII Data on multiple algorithms o Support Vector Machine (SVM) : Traditional Approach o Improved Hybrid Approach for Prediction using Random Forest Approach : Proposed Approach  Implementation of Algorithms and Fetching Results on Multiple Parameters o Execution Time o Cost Factor o Complexity o Performance 3. TOOLS AND TECHNOLOGIES USED FOR IMPLEMENTATION  Python  Notepad++  Scikit-Learn  Python Machine Learning Libraries  SeaBorn  Physionet Polyman EDF Browser 4. IMPORTANCE OF NOVEL APPROACH  Deep Evaluation of the Architecture of EEG Data  Predictive Mining based on the different parameters associated with EEG  Evaluation and Analytics of EEG Aspects with Deep Signal Processing  Analytics of EEG Signals and Waveforms for identification of disorders  Evaluation of Sleep data for different applications  Predictive Analysis from EDF files and association with frames  Electroencephalography (EEG) signals characterization with respect to various states of human body. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 05 Issue: 07 | July 2018 www.irjet.net p-ISSN: 2395-0072