Predicting Type-2 Diabetes Using Machine Learning and Feature Selection Techniques

Abstract

In medical science, the related data for particular disease have high diversities. So finding relationship between those data has been a challenging work. Nowadays machine learning has been experimenting for finding the relationship and pattern between those data and various systems. Diabetes is one of the most common and deadly disease all around the world. This disease has several complexities and can affect various organs of a human body. But till now there is no prevention or cure for diabetes. So predicting that a person might have high chance to develop diabetes can lead a better option. This paper discusses about various approaches for predicting the Type-2 diabetes and also for determining the most important features for developing diabetes. Data are collected from biographical information and pathological test reports of 15000 people. Gradient Boosting, Decision Tree, Random Forest, K-nearest neighbors and Support Vector Machine have been used in a supervised environment to predict the diabetes. Principal Component Analysis (PCA) is applied to find out the important attributes in predicting diabetes. Ensemble method is also applied for dealing with week predictors to get better results.