SVM parameter optimization using swarm intelligence for learning from big data

Abstract

Support vector machine (SVM) is one of the most successful machine learning algorithms to solve practical pattern classification problems. The selection of the kernel function and its parameter plays a vital role on the results. Radius basis function (RBF) is a prevalently used kernel. For an RBF-SVM, two parameters, c and γ, control the SVM performance. In this paper, we present a SVM parameter learning algorithm, DL&BA, effective for learning from big data. The DL&BA algorithm has two stages. At the first stage, we use a distributed learning (DL) to search for a region which promises optimal parameter pairs. At the second stage, a swarm intelligent optimization algorithm - the Bees Algorithm (BA) is used to search for an optimal pair of c and γ. We applied the DL&BA algorithm to solving an important automotive safety problem, driver fatigue detection, which involves a large amount of real-world driving data. Our experimental results show that DL&BA is not only computational efficient but also effective in finding an optimal pair of c and γ.