【摘要】：針對(duì)高維數(shù)據(jù)中不相關(guān)特征、冗余特征等導(dǎo)致的分類任務(wù)計(jì)算量大、分類正確率低等問題,提出了一種基于歸一化互信息的相關(guān)性快速過濾特征選擇(FCBF-NMI)算法.該算法采用歸一化互信息代替對(duì)稱不確定性作為FCBF算法的相關(guān)性評(píng)價(jià)標(biāo)準(zhǔn),進(jìn)行特征與類別、特征與特征的相關(guān)性分析,刪除不相關(guān)特征及冗余特征以獲得最優(yōu)特征子集.實(shí)驗(yàn)結(jié)果表明:FCBF-NMI算法得到的最優(yōu)特征子集更合理,平均分類正確率為89.68%,所用時(shí)間平均低至2.64s.
[Abstract]:A fast feature selection algorithm based on normalized mutual information (FCBF-NMI) is proposed to solve the problems of large computation and low classification accuracy caused by irrelevant features and redundant features in high dimensional data. In this algorithm, normalized mutual information is used instead of symmetric uncertainty as the criterion for evaluating the correlation of FCBF algorithm. The correlation analysis of features and categories, features and features is carried out, and irrelevant features and redundant features are deleted to obtain the optimal feature subset. The experimental results show that the optimal feature subset obtained by the FCBF-NMI algorithm is more reasonable, the average classification accuracy is 89.68 and the average time used is as low as 2.64 s.
【作者單位】： 蘭州理工大學(xué)計(jì)算機(jī)與通信學(xué)院;
【基金】：國家自然科學(xué)基金資助項(xiàng)目(61363078) 甘肅省青年科技基金資助項(xiàng)目(148RJYA001)
【分類號(hào)】：TP301.6
，

本文編號(hào)：2236747