Predicting the Risk of Being a Diabetic Patient Using Statistical Analysis and Data Mining

Abstract

There is a vast and enormous amount of data available in hospitals and medical related institutions. But, the amount of knowledge obtained from such data is very little. Applying IT knowledge for healthcare is an emerging field of huge importance for providing prognosis as well as a deeper understanding of medical data. Diabetics is actually a disease which is affecting many people today Early prediction of diabetes is an extremely challenging task because of the complicated interrelationship between various factors. This research tried to diagnose diabetes which based on 12 risk factors using data from 200 people and applied data mining and statistical techniques to predict the risk of being a diabetic patient. Statistical model has been created using Minitab with the application of the binary logistic regression model. The created model provided the way of predicting the possibility of having diabetics for any person and identified the most suitable risk factors which are most relevant to the disease prediction. Through this identified risk factors, we clustered the data using k-means. An empirical study has proved the effectiveness of our proposed approach.