Classifier fusion for liver function test based Indian jaundice classification

Health information technology is spreading all over the world to provide right medical advice for healthcare. In this context, this paper describes the application of data mining techniques to identify jaundice by analyzing liver function test reports. Artificial neural network and support vector machine classifiers are employed here for classification on the basis of liver condition. Since correct medical advice is a sensitive decision making process, seeking multiple expert opinions may be a good choice. Henceforth, classifier fusion technique is proposed for this work. Two kinds of classifier fusion techniques viz. Decision template and Dempster-Shafer theory are tested here. Several sets of MLP and SVM classifiers are combined by both the fusion techniques. Results reveal that fused multiple classifier gives better prediction accuracy as compared to single classifier. Hybridization of two ANNs and four SVM classifiers with Dempster-Shafer algorithm gives up to 97.33% of prediction accuracy.