Analysis of Depth of Entropy and GINI Index Based Decision Trees for Predicting Diabetes

Abstract

Diabetes is a disease caused due to malfunctioning of pancreas. In this disease, pancreas either no longer produces insulin or produces insufficient insulin. If this disease is diagnosed early, several health complications can be avoided by taking precautions timely. Otherwise, it may create serious health problems. Nowadays, machine learning models are widely researched for diabetes prediction. This research work uses decision tree classifiers for diabetes prediction and analyzed the impact of decision tree depth in diabetes prediction. Besides this, the research work compared the performances of ID3 and CART decision trees in reference to diabetes prediction. From the empirical observation, we concluded that the CART algorithm has slightly better performance than ID3 and the best prediction performance can be achieved with the decision trees of depth 4.