Drug Review-based Diabetes Prediction by Using Naïve Bayes Method

Abstract

Diabetes is a metabolic disease characterized by hyperglycemia caused by defects in insulin secretion, insulin action, or both. Several studies show that late and inappropriate treatment in diabetes mellitus will cause uncontrolled blood glucose in the long term. This condition causes severe changes in heart, brain blood vessels and leg blood vessels, nerves, kidneys, and eyes. Hence, the ability to recognize the existence of diabetes is necessary to prevent the worse condition. This study utilizes the Naive Bayes method to predict diabetes based on drug reviews. The N-Gram and TF-IDF (Term Frequency– Inverse Document Frequency) methods are used for feature extraction. We found that the utilization of the uni-bigram+trigram feature produces the best result with the values of accuracy and F1-score are 0.928 and 0.932, respectively.