Detection of Diabetes using Feature Based Machine Learning Techniques

Abstract

Diabetes is one of the chronic diseases, that trouble the human life worldwide. Diabetic patient found everywhere in the world. Health professionals describes diabetes as a metabolic disease. In this disease a person blood sugar became irregular, due to either inefficient insulin production or a body cells or failure to respond properly to insulin. Glucose level in the blood will rise because of diabetes. Many problems might be face if diabetes remain untreated. The diabetes is not only a disease, it is an originator of many dangerous diseases such as heart attack, kidney failure and blindness. In most of the countries, diabetes has become main cause of illness and death. If we detect diabetes in earlier stage, then it can be controlled. The main goal of proposed work is to early detection of diabetes on the basis of basic symptoms present in the patient. In the proposed model, we applied some machine learning algorithms such as Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbour (KNN) and Random Forest (RF). The performance of every model is measured and compared with other machine learning approaches. The experimental result of proposed model shows that SVM (Linear Kernel) achieved higher accuracy (ACC=0.8037) when dominating attributes were used. When minimum redundancy and maximum relevance(mRMR) were used for dimensionality reduction, logistic regression achieved higher accuracy (ACC=0.8208).