基于PCA鸟群算法的SVM参数优化及应用

摘要

为提高支持向量机的分类性能和寻优速率,研究群体仿生智能算法在参数优化过程中的特点,提出一种基于主成分分析的鸟群算法.通过模拟鸟群的觅食、警觉、迁徙等生物行为,结合主成分分析消除数据之间线性冗余的特点,有效增强模型的泛化能力,降低参数的寻优时间,改善识别精度.采用该算法解决语音识别的参数寻优问题,将其仿真结果与其它算法进行比较,比较结果表明,该算法比标准的鸟群算法和粒子群算法有更快的收敛速度和更高的识别准确率.%To improve the classification performance and optimization rate of support vector machine,characteristics of swarm intelligent optimization on parameter optimization were studied,an optimization method based on principal component analysis of a bird swarm algorithm was proposed.This method effectively enhanced the generalization ability of the model,reduced the optimization time of parameter and increased the recognition accuracy by imitating the biological behavior of birds foraging,vigilance and migration,and combining with the characteristics of principal component analysis to eliminate the linear redundancy between data.The proposed algorithm was used to solve the parameter optimization problem of speech recognition,and the results of simulation were compared to other algorithms.Comparative results indicate that the algorithm has better convergence rate and higher recognition accuracy than simple bird swarm algorithm and particle swarm algorithm.