基于PSO的支持向量机改进算法研究

Research on Improved Algorithm of Support Vector Machine Based on Particle Swarm Optimization Algorithm

提出一种改进的PSO-SVM算法。首先在文本预处理时,针对互信息方法存在的低频特征词倚重,忽略了高频特征词的不足之处,引入了权重因子、类内和类间离散因子进行属性约简;然后基于支持向量机分类模型,以不同核函数相结合构造混合核函数,利用粒子群优化算法(PSO优化算法)迭代寻找全局最优参数组合,并以此参数构造混合核函数结合哈夫曼最优二叉树生成分类器。使用经典数据集对PSO-SVM算法进行性能分析,实验表明改进后的算法可以有效地减少冗余属性,降低计算复杂度并具有更高的准确率和召回率。

In this paper, an improved PSO-SVM algorithm is proposed. Firstly, in the text preprocessing, aiming at the shortcomings of mutual information,relying heavily on the low frequency characteristic words and ignoring the high frequency characteristic words,the weighting factor,the intra-class factor and the inter-class factor are introduced to reduce attributes. Then,based on the Support Vector Machine classification model,the hybrid kernel function is constructed by combining different kernel functions,and the particle swarm optimization algorithm(PSO optimization algorithm) is used to iteratively find the global optimal parameter combination,and the mixed kernel function combined with the optimal binary tree of Huffman is constructed to generate the classifier. The performance of the PSO-SVM algorithm is analyzed by using the classical data set. The experimental results show that the improved algorithm not only effectively reduces the redundant attributes and computational complexity but also has higher accuracy and recall rate.