Diabetes disease prediction using significant attribute selection and classification approach

Abstract

Data Mining performs a major role in healthcare services because disease recognition and investigation contains a vast amount of data. These conditions generate several data managing problems, and to operate efficiently. The healthcare datasets are undefined and influential and it is extremely monotonous to manage and to operate. To get better of the exceeding problems, numerous analyses present various ML algorithms for different disease examination and prediction. The undertaking of disease identification and prediction is an element of classification and forecasting. In this paper, diabetes is estimated by major characteristics and the relation of contradictory characteristics is also categorized. Significant features selection was done via the recursive feature elimination with random forest. The estimation of our system specifies a powerful alliance of diabetes with (BMI) and with glucose level was drawing out using the Apriori approach. XGBoost has examined for the estimation of diabetes. The XGBoost gives better accuracy of 78.91% compared to the ANN approach and might help support medicinal professionals through treatment decisions.