Data Mining and Machine Learning Approaches and Technologies for Diagnosing Diabetes in Women

Abstract

Data mining and machine learning are two interesting areas of computer science that go hand in hand in identifying hidden patterns and extracting valuable information from data. Indeed, Data mining covers the entire process of data analysis including machine learning which aims at constructing programs that learn automatically from experiences. The main purpose of this paper is to make a comparative study of four well-known classification algorithms namely Naive Bayes, Neural network, Support vector machines and Decision tree in order to categorize female patients into two groups; having diabetes or not. Therefore, after adopting well-chosen criteria based on confusion matrix, we run the selected algorithms in two different data mining technologies Weka and Orange. Indeed, the results obtained demonstrate that support vector machines; implemented in Weka toolkit as SMO, is the best technique in terms of accuracy, sensitivity and precision when handling diabetes in women dataset.