Hybrid wavelet ν-support vector machine and chaotic particle swarm optimization for regression estimation

摘要

In view of the bad approximate results of the existing support vector (SV) kernel for series influenced by multi-factors in quadratic continuous integral space, combining wavelet theory with kernel technique, a wavelet kernel function is put forward in quadratic continuous integral space. And then, wavelet ν-support vector machine (W ν-SVM) with wavelet kernel is proposed. To seek the optimal parameters of W ν-SVM, embedded chaotic particle swarm optimization (ECPSO) is also proposed to optimize parameters of W ν-SVM. The results of application in car sale estimation show that the estimation approach based on the W ν-SVM and ECPSO is effective and feasible. Compared with the traditional model, W ν-SVM method requires fewer samples and has better estimating precision.