Type 2 Diabetes Prediction Using Lightgbm

Abstract

Today every second person in the world is suffering from diabetes. Diabetes occurs when the level of glucose abnormally increases in the blood. Among the various types of diabetes Type 2 diabetes is the most common type. The symptoms in type 2 diabetes develop very slow or are absent in the patient. So it becomes almost impossible for the patient to recognize type 2 diabetes in the early stages. Conventional methods take too much time and effort. The proposed system predicts type 2 diabetes in patients using LightGBM algorithm on Pima Indian Dataset. The Pima Indian Dataset has eight features that were chosen to form the basis for forecasting the presence of diabetes within five years in Pima Indian women. Advantages of LightGBM over other algorithms include its faster training speed and efficiency, lower memory usage, better accuracy. To compare the performance of LGBM we used SVM and Random forest algorithms on the same dataset. LightGBM (cv) gives accuracy 0.9062 which is higher in comparison to SVM and Random forest.