InoteveraGaidhi rinotevera
Local Outlier Factor
Tekinoroji
Nhungamiro yehunyanzvi
Multicollinearity inoitika apo vanofanotaura mukudzoreredza vanotakura ruzivo rwakapfuura, zvichiita kuti zviome kuparadzanisa mipiro yavo yega.
Teerera kugwaro iriAI-yakarondedzerwa · ~3 min
The variance inflation factor (VIF) inotara kuti musiyano wecoefficient wakawedzerwa zvakadii nehukama hwayo hwemutsara nevamwe vanofanotaura, asi haisarudze kuti ndezvipi zvinosiyana zviri mumuenzaniso.
A regression coefficient inodudzirwa apo yakabata mamwe mafungidziro akagadziriswa. Kana iyo data iine mashoma mashoma apo imwe fungidziro inoshanduka yakazvimiririra kune imwe, iyo kuenzanisa inotsigirwa zvisina simba. Iyo yakarongedzerwa modhi inogona kuramba ichifanotaura mhinduro zvakanaka, asi shanduko diki kumuenzaniso inogona kuburitsa shanduko huru mune yega coefficients, zviratidzo zvavo, kana zvikanganiso zvakajairwa. Ichi ndicho chikonzero collinearity kazhinji ichikuvadza tsananguro zvakananga kupfuura kufanotaura. For predictor j, idzorere pane mamwe mafungidziro uye uverenge coefficient yekutsunga R-squared. Kusiyana kwayo inflation factor ndeye VIF_j = 1/(1 - R-squared_j). Kana mamwe mafungidziro akatsanangura mazhinji ekufungidzira j, denominator idiki uye VIF yayo yakakura. Muchiverengero chakareruka chekufungidzira, kana R-squared_j iri 0.8, ipapo VIF iri 1/(1 - 0.8) = 5. Pasi pemutsara-muenzaniso setup, izvi zvinoreva kusiyana kweiyo coefficient kashanu izvo zvingave pasi pe orthogonal predictor kuenzanisa nekuenzanisa kwakasara kusiyana; mhosho yaro inopetwa neskweya mudzi weshanu. VIF ndeyekuongorora, kwete yepasirese kupasa / kutadza chikumbaridzo. Hukoshi hukuru hunozivisa kupindirana asi haugone kuziva kana zvine hunyanzvi hwesainzi, kubvisa mutsauko, kana kuti modhi haisiyo. Yakakwana mutsara kutsamira kunoita kuti coefficients asazivikanwe mune yakajairwa dhizaini matrix, nepo pedyo nekutsamira kunosiya fungidziro inogoneka asi isina kugadzikana. Ongorora coefficient kusavimbika, kufanotaura tsananguro, dhizaini matrix uye chinangwa chekudzidza. Mhinduro dzinogona kuitika dzinosanganisira kuunganidza zvakacherechedzwa zvinopatsanura zvinokonzeresa mhedzisiro, kubatanidza matanho akawandisa kana akakodzera, kusarudza imwe fungidziro zvichienderana neruzivo rwekare, kana kushandisa nguva dzose senge ridge regression kana kufanotaura ndicho chinangwa. Kubvisa musiyano chete kudzikisa VIF kunogona kugadzira kusiiwa-inoshanduka bias kana kushandura mubvunzo. Centering inogona kubatsira nekusakosha kuwirirana kwakaunzwa ne polynomial mazwi kana kudyidzana, asi haigadzirise kupindirana kwese kwakakosha. Refit uye ongorora muenzaniso wakasarudzwa, uye tsanangura izvo coefficients ayo anogona uye asingagoni kutsigira.
Iyo yekufungidzira yekugara modhi inosanganisira ese ari maviri nzvimbo yepasi mumakumbo masikweya uye nzvimbo yepasi mune masikweya metres. Nekuti imwe yakagadziriswa rescaling yeimwe, iyo dhizaini matrix haina basa; kuchengetedza imwe unit inodzivirira ruzivo rwakafanana.
Muongorori anofungidzira modhi yenguva yekufamba ine chinhambwe uye inofungidzirwa kushandiswa kwemafuta, izvo zvinonyatso famba panzira dzakatorwa. Iyo yakakwira VIF mireza isina kugadzikana, kunyangwe fungidziro ichiramba ichibatsira mukati menzira dzakafanana.
Muongorori wezvehutano anonyora makore nemakore kubva pakuzvarwa zvakasiyana. Ruzivo rwavo rwakagovaniswa runoita kuti kududzira kwakasiyana-siyana kuve kusina simba; domain reasoning inogona kuona kuti chiyero chimwe kana mutsauko wakasiyana unopindura mubvunzo wechidzidzo.
Chikwata chinofananidza VIF isati yatanga uye mushure mekuisa chigaro chekufungidzira chinoshandiswa mune polynomial modhi. Centering inogona kuderedza kusakosha collinearity pakati pematemu akaomeswa uye ane squared, uku ichichengetedza kukosha kwekuongorora kududzira uye dhizaini yemuenzaniso.
Regression mishumo inogona kuita kuti collinearity ive nyore kutonga nekuratidza tsananguro yekufanotaura, nguva yekuenzanisa, VIF diagnostics uye kufanotaura kuita pamwe chete. Ongororo yeramangwana inofanira kusiyanisa kana chinangwa chiri chakagadzikana chirevo, kufanotaura, kana zvese, sezvo mishonga inosiyana. Kana kuunganidza data idzva kunogoneka, kuwana nemaune makesi anosiyana ekufungidzira akazvimirira anogona kuvandudza kududzira. Kana kuunganidzwa kwedata kusingashanduke, modhi yakajairwa kana chiyero chakajeka chakasanganiswa chingave chakakodzera, nekutengeserana kwayo kwakanyorwa. Dzokorora kuongorora kana chimiro chakasetwa kana huwandu hwevanhu hwachinja; hukama hwezuro hwekufungidzira haufanire kutsanangura sampuli nyowani.
Tsamba yemahara
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Multicollinearity inoitika apo vanofanotaura mukudzoreredza vanotakura ruzivo rwakapfuura, zvichiita kuti zviome kuparadzanisa mipiro yavo yega. The variance inflation factor (VIF) inotara kuti musiyano wecoefficient wakawedzerwa zvakadii nehukama hwayo hwemutsara nevamwe vanofanotaura, asi haisarudze kuti ndezvipi zvinosiyana zviri mumuenzaniso.
Imwe column inogara ichiwanda yeimwe, saka modhi haigone kupatsanura macoefficients maviri.
Collinearity inogona kuita yakaparadzana coefficient fungidziro kusagadzikana kunyangwe fungidziro mukati meiyo inocherechedzwa hutongi hunobatsira.
VIF kuongororwa kunopindirana mumakoramu akabatanidzwa ekugadzira; haritauri pacharo mushonga kana kuongorora kufanotaura.
Kusiyana kweinflation neshanu kunoenderana nestandard-error inflation nesquare root yeshanu.
Iyo ficha inogona kuve yakakosha mune zvesainzi, uye kuibvisa kunogona kuchinja dudziro kana kusiya ruzivo rwakakodzera.
Ramba uchidzidza
Mamwe madhairekitori akasarudzirwa nyaya iyi
InoteveraGaidhi rinotevera
Local Outlier Factor
Tekinoroji