Prognosis of Diabetes Mellitus Based on Machine Learning Algorithms

One of the life-threatening and deep root diseases leading to raise the level of sugar in the blood is Diabetes Mellitus (DM) and if it is kept anonymous and untouched then many difficulties have to be faced as a result. Generally, when it is identified a patient visits a diagnostic center and takes a consultation with the doctor. Now, this critical problem can be solved using Machine Learning (ML) approaches as it is booming. A model has been designed in this study that can prognosticate the prediction of DM in patients with a certain level of accuracy. Hence 3 ML algorithms based on classification namely Decision tree, Support Vector Machine (SVM), and Naïve Bayes are used in our study for early detection of DM. Database which is sourced from UCI ML repository namely Pima Indians Diabetes Database (PIDD) is used for performing experiments. Measurement of the performance of all 3 algorithms is based on various criteria like Accuracy, Recall, Precision, and F- Measure. Measurement of Accuracy is done on both correct and incorrect classified instances. After comparative analysis, it has been found that Naïve Bayes is having the highest accuracy among other algorithms. For verification of results Receiver Operating Characteristic (ROC) curves are traced in an organized manner.