Distributed fuzzy rough prototype selection for Big Data regression

Size and complexity of Big Data requires advances in machine learning algorithms to adequately learn from such data. While distributed shared-nothing architectures (Hadoop/Spark) are becoming increasingly popular to develop such new algorithms, it is quite challenging to adapt existing machine learning algorithms. In this paper, we propose a solution for big data regression, where the aim is to learn the regression model over large high-dimensional datasets. First, a new distributed implementation of the weighted kNN regression method is presented followed by a novel distributed prototype selection method based on fuzzy rough set theory. Experiments demonstrate that our implementations in Apache Spark for the proposed distributed algorithms handle the size and complexity of modern real-world datasets well. We furthermore show that application of our prototype selection method improves the regression accuracy.