Abstract:

Parameter optimization of support vector machines (SVMs) is an important research direction, however, there is a lack of systematic theoretical guidance for SVM parameter selection. The fruit fly optimization algorithm (FOA) can find a better approximate optimal solution quickly and then iterates in the solution neighborhood, but the search time is prolonged. We therefore propose a fruit fly coupled uniform design algorithm (FFUD), which couples the FOA and the uniform design method (UD) to solve the problem of SVM parameter optimization. The proposed algorithm can quickly gain an approximate optimal solution to the problem via the parallel optimization performance of the FOA, and then jumps out of the FOA and continues to perform the local optimization searching through the UD method to obtain a much better approximate optimal solution. Experimental results show that the proposed algorithm has better searching efficiency and higher classification accuracy, and it is effective and feasible in the SVM parameter optimization.

Key words: fruit fly optimization algorithm (FOA);support vector machine (SVM);uniform design (UD);parameter optimization;approximate optimal solution