Diabetes Prediction Using Colab Notebook Based Machine Learning Methods

Abstract

Diabetes is getting more and more common around the world. People suffer from diabetes or live at risk associated with this disease. It is necessary to prevent health problems caused by diabetes, to reduce the risk of diabetes and to reduce the load of diabetes on the health system. Therefore, it is important to diagnose and treat diabetic patients early. In this study, Pima Indian Diabetes (PID) database was used to predict diabetes. The PID database was divided into 2/3 for the training dataset and 1/3 for the test dataset. Then, the test and training datasets were fed into the machine learning models using five-fold cross-validation. Random Forest Classifier, Extra Tree Classifier and Gaussian Process Classifier machine learning methods were used to predict whether individuals have diabetes or not. In this study, the proposed method with the highest prediction accuracy was determined as the Random Forest Classifier. The proposed method's accuracy was 81.71%, precision was 88.79%, recall was 84.83%, F-score was 86.76% and ROC AUC was 88.03%. The proposed method was developed to assist clinicians in predicting the diagnosis of diabetic patients. The machine learning methods developed in this study were applied using Colab Notebook a Google Cloud Computing service.