基于PSO优化的盲源分离式文本特征降维分类方法

The Reduction Dimension Classification Method of Blind Source Separation Text Feature on PSO Optimization

为了有效解决文本特征分类过程中高阶相关性问题,本文在盲源分离式文本特征降维分类方法的基础上引入粒子群(PSO)算法,有效规避迭代过程中局部最优解问题,且以负熵作为适应度函数,有效改善独立主成分分析的判别性能,经过实验证明经过优化后的方案,在精确度、准确率、召回率、F1测试值等方面有较好的表现。

In order to effectively solve the problem of high-order correlation in text feature classification, particle swarm optimization (PSO) algorithm was introduced on the basis of Blind Source Separation (BSS) text feature dimension reduction classification method to effectively avoid the local optimal solution problem in the iteration process. fitness function was regarded as Negative entropy to effectively improve the discriminant performance of independent principal component analysis. Experiments showed that the optimized scheme had better performance in accuracy, accuracy, recall and test value.