Factors Affecting Machine Learning Algorithms Selection

Abstract

Selection of the Machine Learning (ML) algorithms and ML Libraries affect accuracy, response time, scalability and success of implementing new Big Data applications. Unfortunately, algorithms providing high accuracy not necessarily provide good response time and scale well. Different algorithms take different training time and different efforts for operationalization. In this paper we will discuss results of collaborative efforts on benchmarking ML algorithms and libraries and review the algorithm of recommender selecting the appropriate ML algorithm and ML library for new Big Data applications, depending on relative importance of accuracy, response time, scalability and other criteria.