A Machine Learning Technique for Detection of Diabetes Mellitus

Abstract

The need for early detection of diabetes mellitus has led to the development of various intelligent systems using machine learning and artificial intelligence for the recognition of the presence of the disease. However, most of the techniques have yielded a comparatively lower accuracy. This research applied data science techniques to a dataset of diabetes mellitus to improve the accuracy of the prediction of the disease. This was achieved by pre-processing the data with dummy categories and applying principal components analysis for reduced dimensionality. Support vector machine, random forest classifier, and deep neural networks were then used to train the system. Support vector machine, random forest classifier, and deep neural networks yielded accuracies of 0.76, 0.77, and 0.89 respectively. Correspondingly, deep neural networks yielded the highest accuracy. The study concluded that better pre-processing will improve the accuracy of machine learning algorithms in the prediction of diabetes mellitus.