Implementation of the Adaboost Method to Increase the Accuracy of Early Diabetes Predictions to Prevent Death Decision Tree-Based

Abstract

Diabetes is a disease that does not show apparent, immediately visible symptoms, making patients who suffer from it unaware that they have diabetes. Therefore, diabetes is usually only discovered when it has damaged vital parts such as the kidneys, eyes, and human nerves. According to WHO, diabetes is the 9th most deadly disease in the world. Early predictions for diabetes are needed based on the supporting attributes of diabetes. In writing this research, a decision tree algorithm method was used. This algorithm still needs to improve in making accurate predictions. So, we need a way to increase the accuracy of data mining learning results. This research aims to improve prediction learning results for diabetes by using an ensemble learning method, namely Adaptive Boosting (AdaBoost). This method was tested by predicting whether someone had diabetes or not. This research method uses CRISP-DM, and the data used is from the Kaggle dataset of 520 data records. The research results showed that the decision tree algorithm produced an accuracy of 94.23%, and the decision tree algorithm added with Adaboost had an accuracy of 97.31%