Machine learning on medical dataset

Abstract

Machine learning for medical decision-making is helpful when it is possible to achieve two conditions: a well-defined way to reach the diagnosis and high prediction accuracy. With the introduction of sensors in every aspect of daily live, there is a huge amount of data compiled (the present study concentrates on data from health), such quantity without any processing/information extraction technique means nothing. However, not every processing makes it possible to obtain something helpful from the raw data. In the present, the document evaluated three methodologies about machine learning (ANN + PSO, SVM, and RVM) in order to make a valuation on the performance over a set of data, and after the tests, the RVM methodology behaves the best.