ISSN (Online): 2349-7084 GLOBAL IMPACT FACTOR 0.238 DIIF 0.876 INTERNATIONAL JOURNAL OF COMPUTER ENGINEERING IN RESEARCH TRENDS VOLUME 2, ISSUE 7, JULY 2015, PP 437-443 IJCERT © 2015 Page | 437 http://www.ijcert.org An Expert System based on SVM and Hybrid GA-SA Optimization for Hepatitis Diagnosis S. Anto, S. Chandramathi Abstract— An accurate diagnosis of diseases like hepatitis is a challenging task for physicians. This problem in diagnosis has attracted researchers to design medical expert systems with utmost accuracy. This paper proposes a clinical decision support system based on Support Vector Machine (SVM) and hybrid Genetic Algorithm (GA) –Simulated Annealing (SA) for the diagnosis of hepatitis by using the dataset of UCI machine learning repository. The SVM with Gaussian Radial Basis Function (RBF) kernel performs the classification process. The hybrid GA-SA is used for two purposes, one is to select the most significant feature subset of the dataset, and the other is to optimize the kernel parameters of SVM. The performance of the expert system is analyzed using various parameters like classification accuracy, sensitivity and specificity. The classification accuracy of the proposed system is found to be superior to that of the other existing systems in the literature. Index Terms— Medical Expert System, Machine Learning, Genetic Algorithm, Simulated Annealing, Support Vector Machine. ——————————  —————————— 1 INTRODUCTION achine learning focuses on the improvement of machine projects that develop and change when given new information. It is of two types: Supervised learning and Unsupervised learning. Supervised learning is the task of machine learning that infers a function from labeled training data. Unsupervised learning is a type of machine learning algorithm used to draw inferences from datasets consisting of input data without labeled responses. There are many examples of machine learning problems such as Optical Character Recognition (OCR), face detection, medical diagnosis and weather prediction. For decision making in data mining, classification is the most common technique used. They help users in understanding how the category relates to other categories. Utilization of the classification system in medical diagnosis had drawn more attention. Assessment of information taken from patients and decisions of experts are the foundations of diagnosis. Determination of best features has more effect on the precision of the analysis framework in prediction. An automatic diagnosis problem can be approached via both single and hybrid machine learning methods. The proposed work diagnoses hepatitis disease. Viral hepatitis is a significant health issue around the globe [1]. Hepatitis is the aggravation and damage to hepatocytes in the liver and can be caused by autoimmunity, viruses, infections with fungi and bacteria, or exposure to toxins such as alcohol. Hepatitis diseases can be spread through blood transfusion and shared syringes. ————————————————  S.Anto is currently working as Assistant professor in Dept. of Computer Science and Engineering , Sri Krishna College of Technology, Coimbatore, India .E-mail : georgeantocse@gmail.com  Dr.S.Chandramathi is working as Dean-Electrical Sciences in Sri Krishna College of Technology, Coimbatore, India. E-mail : chandrasrajan@gmail.com 2 RELATED WORKS Seera et al [2] have proposed a hybrid intelligent system for medical data classification. A hybrid intelligent system that consists of the Fuzzy Min-Max neural network, the classification and regression tree, and the random forest model is proposed. The system yields good generalization performance but is abysmally slow in test phase. The accuracy of the proposed system is 78.39%. Calisir et al [3] have proposed an intelligent hepatitis diagnosis system using Principal Component Analysis (PCA) and Least Square Support Vector Machine (LSSVM).The simplest invariance could not be captured, unless the training data explicitly provides this information. Kaya et al [4] have proposed a hybrid decision support system based on Rough Set (RS) and Extreme Learning Machine (ELM) for the diagnosis of hepatitis. Redundant features are removed from the dataset through RS approach and classification process is implemented through ELM by using the remaining features. The performance of the proposed system is decreased due to time delay. Ruxandra Stoean et al [5] have proposed medical decision making model using SVMs, explained by rules of Evolutionary Algorithms (EA) with feature selection.SVMs successfully achieve high prediction accuracy due to a kernel-based engine and EAs can greatly accomplish a good explanation of how the diagnosis was reached. Similarly SVM concept is used for various datasets like diabetes, breast cancer and hepatitis as in literatures [13], [14], [15], [19], [20]. This paper proposes a medical decision support system which uses SVM for classification and GA- SA for optimization as shown in Fig.1.SVM is a kernel based statistical classification technique which is widely used to solve bi-class problems .In this work the SVM uses Gaussian kernel function. GA is an evolutionary algorithm which offers multi criterion optimization for higher dimensional pace problems. It is a popular local search method used for M