Komugqa ekuhlehleni

Regression ukuhlaziywa kungenziwa wanezela izindlela lezibalo ngitadisha ubuhlobo obuphakathi eziguquguqukayo ethize (ezincikile nezizimele). Kulokhu, eziguquguqukayo ezimele abizwa ngokuthi "covariates" ethembela - "criterial". Lapho uqhuba komugqa ekuhlehleni ukuhlaziywa ukumelwa variable engaphansi kuthatha ngesimo esikalini isikhawu. Kukhona amathuba khona ubudlelwano non-eqondile emkhatsini wetinombolo ezihlobene isikali isikhawu, kodwa le nkinga isivele ixazululwe by izindlela ekuhlehleni non-eqondile, engesiyo isihloko lesi sihloko.

Ukuthuthukiswa ekuhlehleni yasetshenziselwa ngempela ngempumelelo njengoba e izibalo, futhi izifundo kwezomnotho esekelwe idatha kwezibalo.

Ngakho cabanga ngalo kwaba nokushiywa ngaphezulu. Ngokombono walaba indlela zezibalo yokunquma ubuhlobo komugqa phakathi ezinye eziguquguqukayo ekuhlehleni komugqa kungenziwa imelelwa ifomula: y = a Bx +. Ukuze uthole incazelo yokuthi lokhu ifomula zingatholakala kunoma iyiphi amabhuku okufunda ngesikhathi Econometrics.

Lapho ekwandiseni isibalo observation (kufika ku Inombolo n-th izikhathi) etholwe kwaba nokushiywa elula eqondile, emelelwa ifomula:

yi = A + bxi + ei,

lapho ei - ezimele, ncamashí zisakazwe, eziguquguqukayo okungahleliwe.

Kulesi sihloko Ngithanda ngonanzelelo lo mqondo ngokombono ukubikezela intengo ngekusasa elisekelwe idatha odlule. Kule ndawo, silinganisela a ekuhlehleni komugqa ngenkuthalo usebenzisa i-okungenani indlela yezikwele, esiza ukwakha umugqa oqondile "ofanele kakhulu" ngokusebenzisa inombolo ethile yamanani yentengo amaphuzu. Idatha okokufaka esetshenziswa iphuzu intengo, okusho okusezingeni eliphezulu, ongaphakeme, ukuvalwa noma ukuvulwa, futhi isilinganiso lezi zimiso (isib isamba esiphezulu kanye esincane ehlukaniswa amabili). Futhi, lezi idatha ngaphambi ukwakha umugqa ofanele kungenziwa ezinqumela eshelelayo.

Njengoba kushiwo ngenhla, ekuhlehleni komugqa ngokuvamile esetshenziswa abahlaziyi ukunquma umkhuba ngesisekelo intengo nesikhathi. Kulokhu, emthambekeni inkomba ekuhlehleni eyonquma ubukhulu izinguquko intengo iyunithi isikhathi ngasinye. Esinye sezimo ngesinqumo lesifanele usebenzisa le nkomba ukusetshenziswa generator isignali, kulandela mkhuba ukuthambekela ekuhlehleni. Uma eyehlelayo omuhle (ekuphumeni komugqa ekuhlehleni) ukuthenga wenziwa uma inani inkomba mkhulu kunoziro. Phakathi emthambekeni ezimbi (nokuncipha ekuhlehleni) ukudayiswa kufanele ibe abakuzuzile inkomba (ngaphansi kuka-zero) ezingezinhle.

Njengoba lisetshenziswe ekunqumeni umugqa engcono elihambisana inombolo ethile intengo amaphuzu, indlela okungenani-izikwele kusho ukuthi i-algorithm ezilandelayo:

- uwukubonakaliswa ingqikithi umehluko izigcawu amanani kanye umugqa ekuhlehleni;

- iyona ratio madlana futhi inani libhala uhla ekuhlehleni idatha uchungechunge;

- phezu yi ngekhompyutha impande square, okuhambelana emgomeni.

Simple Linear Regression sesibalo has imodeli:

y (x) = f (x) ^,

lapho - izici elikhiqizayo wethule variable engaphansi;

x - achazayo noma variable ezimele;

^ Ikhomba ukungabikho esiqinile ubuhlobo obusebenzayo phakathi eziguquguqukayo x kanye y. Ngakho-ke, ngamunye endabeni ethile, y variable kungase kuhilele lafana:

y = yx + ε,

lapho - umphumela idatha langempela;

uh - yi idatha theory kunqunywa ngokuxazulula ezothando ekuhlehleni ;

ε - kwenombolo engahleliwe okuyinto ephawula ukuphambuka phakathi ukubaluleka langempela kanye theory.