基于條件信息熵的超高維分類數(shù)據(jù)特征篩選
發(fā)布時間:2019-02-14 19:02
【摘要】:文章提出了一種基于條件信息熵的超高維自由模型下非參數(shù)特征篩選方法,在響應(yīng)變量為兩類別時,對多類別離散型協(xié)變量進(jìn)行特征篩選。通過理論證明和蒙特卡羅數(shù)值模擬驗證了該篩選方法具有確定篩選性質(zhì),對超高維分類變量的重要特征篩選具有顯著的效果。
[Abstract]:In this paper, a non-parametric feature selection method based on conditional information entropy for ultra-high dimensional free model is proposed. When the response variable is two classes, the multi-class discrete covariable is selected. The theoretical proof and Monte Carlo numerical simulation show that the method has definite screening properties and has remarkable effect on the important feature screening of ultra-high dimensional classification variables.
【作者單位】: 南京信息工程大學(xué)數(shù)學(xué)與統(tǒng)計學(xué)院
【基金】:國家自然科學(xué)基金資助項目(11301279) 江蘇省自然科學(xué)基金資助項目(BK20140983;BK20161530) 江蘇省“青藍(lán)工程”資助項目(2016)
【分類號】:O21;O236
本文編號:2422501
[Abstract]:In this paper, a non-parametric feature selection method based on conditional information entropy for ultra-high dimensional free model is proposed. When the response variable is two classes, the multi-class discrete covariable is selected. The theoretical proof and Monte Carlo numerical simulation show that the method has definite screening properties and has remarkable effect on the important feature screening of ultra-high dimensional classification variables.
【作者單位】: 南京信息工程大學(xué)數(shù)學(xué)與統(tǒng)計學(xué)院
【基金】:國家自然科學(xué)基金資助項目(11301279) 江蘇省自然科學(xué)基金資助項目(BK20140983;BK20161530) 江蘇省“青藍(lán)工程”資助項目(2016)
【分類號】:O21;O236
【相似文獻(xiàn)】
相關(guān)碩士學(xué)位論文 前1條
1 孫超男;基于條件信息熵的超高維分類數(shù)據(jù)特征篩選[D];南京信息工程大學(xué);2017年
,本文編號:2422501
本文鏈接:http://sikaile.net/kejilunwen/yysx/2422501.html
最近更新
教材專著