基于改进蚁群算法的高精度文本特征选择方法

为了反映特征子集对分类结果的整体影响,去除噪声特征项,提出了一种基于改进蚁群算法的高精度文本特征选择方法。建立了特征选择数学模型,实现了特征选择过程与分类器分类过程间的直接关联;设计了特征优选与特征精选相结合的模型求解方案,降低了模型求解过程中的计算复杂度;提出了基于等效路径增强和局部搜索更新相结合的改进蚁群算法,提高了解的质量和稳定性。实验结果表明,与现有文本特征选择方法相比,该方法能大幅提升分类精度。

High precision method for text feature selection based onimproved ant colony optimization algorithm

To reflect the overall impact of feature subset on the classificat ion resul t and remov e the noisefeatur es, a new high-precision method was proposed for text featur e select ion based on the improv ed antco lony opt imization alg orithm. A mathematical model for feature select ion w as established to realize thedirect correlat io n betw een the feature select ion pro cess and the classif ier classificat ion pro cess. A newmodel-solv ing metho d composed of the opt imized feature select ion step and the refined feature selectionstep w as designed and the computat ional complex ity in the pro cess of model solving reduced. A newimproved ant colo ny opt imizat ion algor ithm based o n equivalent routes and local sear ch w hich improv ed thequality and stability of the problem solut io n was proposed. The experiment result s on the tw o dataset sshow the superiority of the pro posed method ov er the current feature select ion metho ds in terms ofclassif icat io n accuracy.