Predicting the Early Stage of Diabetes and Finding the Association of the Symptoms

Abstract

Diabetes has become a growing global public health issue. This illness can become chronic since there are no early symptoms and it can lead to several negative health impacts. This study focuses on the early identification of diabetes based on the symptoms of the disease and finds the relation between the symptoms and the diagnosis. Successful early diagnosis of this disease could boost a person to a better treatment plan before it worsens the health to a critical stage. This study exploits recent classification algorithms including Naïve Bayes, Logistic Regression, REPTree, J48, and Random Forest. The association rules mining using the apriori algorithm is used. Further, the 10-fold cross-validation with split criteria was adopted. The authors adopted several evaluation metrics including accuracy, precision, recall, F-measure, and area under the ROC curve. The research findings revealed the Random Forest to be the best classification algorithm as compared to other classifiers.