PREDICTION OF DIABETES USING MACHINE LEARNING

Diabetes is observed to be among the most perilous diseases and chronic diseases. A multitude of problems arise if the disease is left unattended and untreated. The prosaic task of identification of the problem aggregates to a patient visiting a doctor in a medical center for deliberation. However,the rise of machine learning methodologies solve this severe problem. The incentive of this study is to examine the model which can prefigure the plausibility of diabetes in patients with maximal accuracy. Thereupon, the four algorithms namely Decision Tree, Random Forest, Naive Bayes and Adaboost Classifier are utilized in this research for predicting diabetes at an initial stage. This paper aims at testing befitting algorithms for the prediction of diabetes. Experiments are conducted on two datasets namely Pima India Diabetes Database(PIDD) which is referenced from UCI machine learning repository and an auxiliary database. The efficiency of all four algorithms are assessed on the basis of Accuracy, Precision, Recall and F-Measure. Using these all four algorithms discussed above the result that was acquired reveal Adaboost Classifier exceeds the other algorithms with the highest accuracy of 85.5% for PIDD and 95.4% for the aforementioned dataset. The result that was obtained using Receiver Operating Characteristic (ROC) curves in a sequential manner.The efficie