Evaluation of predictive Machine Learning techniques as expert systems in medical diagnosis

Abstract

Background/Objectives: Medical science industry has immense measure of information; however a large portion of this information is not mined. Machine Learning takes analytics to the extreme by exploring hidden information in data. Disease diagnosis is major intention of medical decision support system which will assist the physicians to obtain valuable decision. Methods/Statistical Analysis: This research work Machine Learning techniques: K-Nearest Neighbors, Decision Tree, Artificial neural networks, Radial Basis Function neural networks and Support Vector Machine are analyzed. Findings: Performance of these techniques is compared through various performance measures such as sensitivity, specificity, accuracy, F measure, and Kappa statistics, True Positive Rate, False Positive Rate and ROC on Breast Cancer Wisconsin, Liver Disorder, Hepatitis and cardiovascular Cleveland Heart disease datasets. Research work consists of 10V fold cross validation method to evaluate the fair estimate of prediction techniques. Application/Improvements: Evaluation of these techniques on diverse medical datasets gave an insight into predictive ability of Machine Learning in medical diagnosis and there is a wide space of improvement.