Performance Evaluation of Supervised Machine Learning Algorithms Using Different Data Set Sizes for Diabetes Prediction

摘要

Data classification algorithm in machine learning is very helpful in analyzing several medical data with a large size and helps in making decisions to diagnose a disease. Not all supervised classification algorithms get accurate results in analyzing data sets. For this reason, testing the accuracy of each supervised classification algorithm is necessary, this can be used as a comparison in determining which types of algorithms are most accurate in measuring small amounts of data, and which algorithms are the most accurate in measuring large amounts of data. In this paper we will examine several classification algorithms including Naïve Bayes algorithms, functions (Support Vector Classifier algorithms), rules (decision table algorithms), trees (J48) by looking at the results of measurements made by each algorithm with measurement variables, which are Correctly Classified, incorrect classifieds, Precision, and Recall. The purpose of the study was to find the weaknesses and strengths of the supervised classification algorithm based on the measurement variables that have been determined against the testing of predictive databases of diabetes. Based on the results in this study, the best algorithm that can be used to help decide to diagnose a disease is the SVM algorithm with an accuracy value of 77.3%.