Abstract:

There is little mathematical theory guidance for the parameter optimization of support vector machines (SVMs), and the traditional artificial bee colony (ABC) is easy to fall into the longterm stagnation. Since the chaotic search algorithm has good randomicity and ergodicity, we propose a parameter optimization model  based on the ABC algorithm with the chaos update strategy (IABCSVM) to solve this problem. This model uses the chaotic search algorithm to improve the searching way of reconnaissance peak, and improve the ABC’s searching efficiency. We evaluate the proposed algorithm on the public data sets from University of California Irvine (UCI), and compare it with the ACOSVM, PSOSVM, and ABCSVM models. Experimental results show that the IABC algorithm is feasible and effective for optimizing SVM parameters, and has higher prediction accuracy and better stability.

Key words: support vector machine, parameter optimization, artificial bee colony algorithm, chaotic search, prediction accuracy