Selecting critical features for biomedical data classification

Abstract

In this paper, the application of machine learning methods on large data sets with numerous features was investigated, with a focus on the identification of critical features in order to reduce the data and produce more accurate results. The research discusses feature extraction techniques for classifying two biomedical data sets with 62 and 71 features, respectively. The results were compared and presented using four classification techniques. The acquired results demonstrate that the selected important features typically produce more accurate results, or at least the same results while reducing the size of the data set and making data collecting easier.