王軍濤

摘 要： 傳統(tǒng)二階特征估計(jì)法在對(duì)大數(shù)據(jù)方差進(jìn)行估計(jì)，預(yù)測(cè)大型教學(xué)系統(tǒng)中的智能大數(shù)據(jù)關(guān)鍵特征時(shí)，存在對(duì)多特征的智能大數(shù)據(jù)關(guān)鍵特征估計(jì)效果不明顯，估計(jì)結(jié)果誤差累計(jì)量大的問(wèn)題。因此，提出大型教學(xué)系統(tǒng)的智能大數(shù)據(jù)關(guān)鍵特征估計(jì)方法，其采用Relief關(guān)鍵特征估計(jì)方法獲取大數(shù)據(jù)特征權(quán)重，完成智能大數(shù)據(jù)特征流行學(xué)習(xí)，通過(guò)對(duì)特征權(quán)重選擇后的數(shù)據(jù)空間進(jìn)行無(wú)監(jiān)督學(xué)習(xí)和低維嵌入，實(shí)現(xiàn)對(duì)多特征的智慧大數(shù)據(jù)的特征估計(jì)?；诖髷?shù)據(jù)關(guān)鍵特征估計(jì)結(jié)果，采用滾動(dòng)時(shí)間序列估計(jì)方法，通過(guò)[AR（p）]模型運(yùn)算大數(shù)據(jù)特征的模型階數(shù)，依據(jù)該階數(shù)向滾動(dòng)AR算法引入實(shí)時(shí)數(shù)據(jù)，解決大數(shù)據(jù)特征估計(jì)中估計(jì)結(jié)果不同步造成的累計(jì)誤差問(wèn)題，實(shí)現(xiàn)智能大數(shù)據(jù)關(guān)鍵特征準(zhǔn)確預(yù)測(cè)。實(shí)驗(yàn)結(jié)果表明，所提方法可增強(qiáng)對(duì)關(guān)鍵特征的估計(jì)精度，對(duì)關(guān)鍵特征的估計(jì)效果也有所提高。

關(guān)鍵詞： 大型教學(xué)系統(tǒng); 智能大數(shù)據(jù); 關(guān)鍵特征; Relief; 時(shí)間序列估計(jì); 累計(jì)誤差

中圖分類號(hào)： TN911?34; TP301 文獻(xiàn)標(biāo)識(shí)碼： A 文章編號(hào)： 1004?373X（2018）12?0083?04

Abstract： The traditional two?order feature estimation method has the problems of unobvious key feature evaluation effect of multi?feature intelligent big data and big error accumulation quantity of evaluation results when it is used to estimate the variance of big data and predict the key features of intelligent big data in the large?scale teaching system. Therefore， a key feature estimation method for intelligent big data in the large?scale teaching system is proposed. The weights of big data features are obtained by using the key feature estimation method Relief to accomplish the popular learning of intelligent big data features. The unsupervised learning and low?dimensional embedding are performed for data space after feature weight selection， so as to realize the feature estimation of multi?feature intelligent big data. On the basis of the key feature estimation results of big data， the model order of big data features is calculated by using the rolling time series estimation method and [AR（p）] model. According to the order， real?time data is introduced to the rolling AR algorithm to resolve the accumulated error problem caused by unsynchronization of evaluation results in big data feature evaluation， so that accurate key feature prediction of intelligent big data can be realized. The experimental results show that the proposed method can improve the estimation precision and effect of key features.

Keywords： large scale teaching system; intelligent big data; key feature; Relief; time series estimation; accumulated error

教學(xué)系統(tǒng)中包含許多智能的大數(shù)據(jù)，如何對(duì)其中關(guān)鍵的特征進(jìn)行準(zhǔn)確估計(jì)成為目前研究的熱點(diǎn)之一，專家和學(xué)者根據(jù)不同教學(xué)系統(tǒng)的數(shù)據(jù)特點(diǎn)已經(jīng)有一些研究成果[1]，但研究還處于初級(jí)階段，傳統(tǒng)二階特征估計(jì)法在對(duì)大型教學(xué)系統(tǒng)中的智能大數(shù)據(jù)關(guān)鍵特征估計(jì)時(shí)，存在特征估計(jì)效果不明顯、特征估計(jì)誤差累計(jì)量大的問(wèn)題。因此，本文研究大型教育系統(tǒng)的智能大數(shù)據(jù)關(guān)鍵特征估計(jì)方法，來(lái)提高關(guān)鍵特征估計(jì)結(jié)果的精度和效果。

1 智能大數(shù)據(jù)關(guān)鍵特征估計(jì)方法

1.1 Relief關(guān)鍵特征估計(jì)方法

針對(duì)大型教學(xué)系統(tǒng)中的智能大數(shù)據(jù)，采取Relief特征估計(jì)方法對(duì)教學(xué)系統(tǒng)中的智能大數(shù)據(jù)的關(guān)鍵特征的權(quán)重進(jìn)行估計(jì)[2]，Relief方法用于數(shù)據(jù)關(guān)鍵特征的估計(jì)是因?yàn)槠淇梢詸z測(cè)一些在統(tǒng)計(jì)上與目標(biāo)屬性不相關(guān)的關(guān)鍵特征。

3 結(jié) 論

本文提出的大型教學(xué)系統(tǒng)的智能大數(shù)據(jù)關(guān)鍵特征估計(jì)方法，可有效提高智能大數(shù)據(jù)的關(guān)鍵特征估計(jì)精度，增強(qiáng)特征估計(jì)效果。

參考文獻(xiàn)

[1] SU H， YANG H， ZHENG S. The large?scale crowd density estimation based on effective region feature extraction method [J]. Lecture notes in computer science， 2010， 6494： 302?313.

[2] 張凱鋒，楊國(guó)強(qiáng)，陳漢一，等.基于數(shù)據(jù)特征提取的風(fēng)電功率預(yù)測(cè)誤差估計(jì)方法[J].電力系統(tǒng)自動(dòng)化，2014，38（16）：22?27.

ZHANG Kaifeng， YANG Guoqiang， CHEN Hanyi， et al. An estimation method for wind power forecast errors based on numerical feature extraction [J]. Automation of electric power systems， 2014， 38（16）： 22?27.

[3] 李炳乾.對(duì)等P2P網(wǎng)絡(luò)中大數(shù)據(jù)關(guān)鍵特征挖掘模型仿真[J].計(jì)算機(jī)仿真，2014，31（11）：294?296.

LI Bingqian. Model simulation of big data key characteristics mining for peer?to?peer networks [J]. Computer simulation， 2014， 31（11）： 294?296.

[4] 張招亮，陳海明，黃庭培，等.無(wú)線網(wǎng)絡(luò)的差異化比特錯(cuò)誤率估計(jì)方法[J].計(jì)算機(jī)研究與發(fā)展，2014，51（1）：138?150.

ZHANG Zhaoliang， CHEN Haiming， HUANG Tingpei， et al. Differentiated bit error rate estimation for wireless networks [J]. Journal of computer research and development， 2014， 51（1）： 138?150.

[5] 樊源泉，伍衛(wèi)國(guó)，許云龍，等.MapReduce環(huán)境中的性能特征能耗估計(jì)方法[J].西安交通大學(xué)學(xué)報(bào)，2015，49（2）：14?19.

FAN Yuanquan， WU Weiguo， XU Yunlong， et al. A power estimation method based on performance features in MapReduce environments [J]. Journal of Xian Jiaotong University， 2015， 49（2）： 14?19.

[6] 梁勝杰，張志華，高紹忠，等.基于變點(diǎn)分析思想的高維機(jī)械噪聲數(shù)據(jù)本征維估計(jì)方法[J].船舶力學(xué)，2016，20（11）：1485?1493.

LIANG Shengjie， ZHANG Zhihua， GAO Shaozhong， et al. A method of latent characteristic variables dimensionality evaluating of high?dimensional mechanical noise data based change?point analysis [J]. Journal of ship mechanics， 2016， 20（11）： 1485?1493.

[7] 唐建生，靳云姬，江向東.目標(biāo)機(jī)動(dòng)時(shí)刻估計(jì)的聲特征變化顯著性檢驗(yàn)方法[J].系統(tǒng)工程與電子技術(shù)，2016，38（2）：270?273.

TANG Jiansheng， JIN Yunji， JIANG Xiangdong. Target maneuver onset time estimation via significance test of acoustic feature variation [J]. Systems engineering and electronics， 2016， 38（2）： 270?273.

[8] BI T， NI Y， SHEN C M， et al. An on?line distributed intelligent fault section estimation system for large?scale power networks [J]. Electric power systems research， 2002， 62（3）： 173?182.

[9] 徐子偉，王鵬，陳宗海.一種基于Fisher比率和預(yù)測(cè)風(fēng)險(xiǎn)準(zhǔn)則的電信客戶流失預(yù)測(cè)分步特征選擇方法[J].中國(guó)科學(xué)技術(shù)大學(xué)學(xué)報(bào)，2017，47（8）：686?694.

XU Ziwei， WANG Peng， CHEN Zonghai. A two?stage feature selection method based on Fisher′s ratio and prediction risk for telecom customer churn prediction [J]. Journal of University of Science and Technology of China， 2017， 47（8）： 686?694.

[10] SHOJI S， KOYAMA A. A fast classification method of isomorphic polygons in design data of large scale integrated circuit [C]// Proceedings of 10th International Conference on Complex， Intelligent， and Software Intensive Systems. Fukuoka： IEEE， 2016： 204?209.