DATA SCIENCE

Abstract

. One of the diseases that is generally characterized by symptoms of an increase in glucose levels in the blood and is one of the body diseases classified as chronic is diabetes. Diabetes suffered by a person from time to time can cause serious damage to other organs such as blood vessels, kidneys, heart and nerves. Machine learning provides various data mining algorithms that can be used to assist medical experts. The accuracy of machine learning algorithms is a measure of the effectiveness of decision support systems. Prediction of diabetes can be seen from the patient's medical record data, therefore the author wants to create a diabetes prediction system independently through a website-based application system. This application system will be combined with data observation, namely the science of data mining using the XGBoost algorithm. The dataset is divided into training data by 80% and testing data by 20%. Before creating the model, various parameter setting scenarios were carried out to evaluate several adjusted parameters, namely colsample_bytree, gamma, learning_rate, max_depth, n_estimators, reg_alpha, reg_lambda, and subsample. After sharing data and tuning parameters, the model results from the XGBoost algorithm have an accuracy of 74.67%, a precision value of 57.40%, a recall value of 65.94% and a specificity value of 78.50%.