State-of-the-art Machine Learning Classifiers: A Comparative Study on health data

Nowadays data mining has gained much eminence. It is a great source of information retrieval. Departments have vast collection of versatile data. It can be in the form of text, digits or images available in online data repositories and data ware houses. In medical field data is related to disease diagnosis, patient's recommendation and different medical situations. Machine learning technique is very helpful in data mining. Machine learning techniques help to analyze Health datasets that can provide facilitation in disease diagnosis and decision making. Previous comparative research work on the machine learning techniques is mostly limited to a single dataset evaluation with few parameters and mostly deficit of statistic performance metrics graphical presentation. The purpose of this research work is to propose and conduct an empirical analysis of multiple widely used machine learning classifiers on health datasets through accuracy, sensitivity, specificity and time parameters measurements. Naive Bayes, LMT, SMO, JRip and J48 Algorithm are used to empirically analyze the Kidney, Diabetes and Liver datasets. The empirical analysis of five classifiers with numeric metrics and performance graphics results that J48 classifier gave best performance on three datasets with different cross validation ratios as compared to other algorithms.