Linear nokushiywa

Uhlalutyo ukuhlehla zingadityaniswa kwi iindlela lwesibalo ukufunda ulwalamano phakathi kwezinto ezithile (kuxhomekeke kunye ezizimeleyo). Kulo mzekelo, eziguquguqukayo ezizimeleyo kuthiwa "covariates" kwaye uxhomekeke - "criterial". Xa uqhuba uhlalutyo yabaxwa ukumelwa oxhomekeke eziguquguqukayo kuthatha uhlobo ngendlela lokuphumla. Kukho amathuba lobukho ubudlelane non-linear phakathi kwezinto ezinxulumene isikali lekhefu, kodwa le ngxaki sele kusonjululwe iindlela parameter nganye non-linear, ingeyo ngumxholo kweli nqaku.

Linear ukuhlehla sasetyenziswa ngempumelelo njengokuba izibalo, kwaye izifundo zoqoqosho ngokususela kwidatha manani.

Ngoko cinga le nokushiywa ngakumbi. Kweli qela indlela zezibalo yokumisela ubudlelwane yomgama phakathi kwezinye izinto yabaxwa angamelwa njenge ifomula: y = a 'bx' +. Ukuze ufumane inkcazelo le fomyula inokufumaneka nayiphi na incwadi kwi oqosho.

Xa nokwandisa inani komkhanya (ukuya kwinani n-th amaxesha) efunyenwe nokushiywa elula yomgama, amelwe ifomula:

yi = A + bxi + ei,

apho ei - ezizimeleyo, ukukhetha isasazwe, variables random.

Kweli nqaku Ndingathanda ukuhlawula ingqalelo ngakumbi kule ingqiqo ngokwembono okubonisa ixabiso elizayo ngokususela kwidatha ezidlulileyo. Kule ndawo, sithekelela yabaxwa eli usebenzisa izikwere indlela ubuncinane, nto leyo eza kunceda ekwakheni umgca ngqo "ifanelekileyo" ngokusebenzisa inani elithile amaxabiso amanqaku kwamaxabiso. Igalelo data esetyenziswa yi kwindawo ixabiso, intsingiselo phezulu, phantsi, ukuvalwa okanye ukuyivula, kwaye umyinge wala maxabiso (umz, wamcacisela inani kwelona qondo umahlule ngo amabini). Kwakhona, ezi iinkcukacha phambi kokuba ukwakha umgca efanelekileyo nga lokungavinjwa walungisa.

Njengoko kukhankanyiwe ngasentla, yabaxwa isoloko isetyenziswa yi ngabahlahleli ukumisela intsingiselo ngenxa nexabiso kunye nexesha. Kulo mzekelo, kwithambeka Isalathisi parameter nganye iya kuqinisekisa ubukhulu lotshintsho ixabiso unit ixesha nganye. Omnye yemiqathango ukuba isigqibo esichanekileyo usebenzisa le salathisi ukusetyenziswa umqondiso wenkunkuma, emva isimo ukuthambekela parameter nganye. Ukuba (ukunyuka yomgama ukuhlehla) ukuthenga kwithambeka entle lwenziwa ukuba ixabiso Isalathisi mkhulu kuno-zero. Ngexesha kwithambeka ezimbi (nokunciphisa ukuhlehla) ukuthengiswa kufuneka ibe amaxabiso akhabanayo le Isalathisi (ngaphantsi kwe zero).

Njengoko kusetyenziswa ekugqibeni umgca kakuhle ehambelana inani elithile lamanqaku ixabiso, indlela ubuncinane-izikwere kuthetha ukuba algorithm ilandelayo:

- lo ibinzana lilonke umahluko wezikweri amaxabiso kunye parameter le;

- lulwahlulo kwale mali kunye nenani imivalo kwi uluhlu lwe parameter nganye data series;

- phezu isiphumo ayinakubalwa ruthi, okuhambelana ecaleni.

Simple Linear nokushiywa Isibalo has imodeli:

y (x) = f (x) ^;

apho - iimpawu nemveliso thaca variable kwayo;

x - achazayo okanye umahluko elizimeleyo;

^ Ibonakalisa ukungabikho ngqongqo ubudlelwane esebenzayo phakathi yezahlukileyo u x no y. Ngoko ke, kwimeko ethile nganye, lo y bume kuqulatha amagama ezinjalo:

y = yx + ε,

apho - idatha ngenxa ngokoqobo;

uh - data japan ithiyori ezimiselwe ekusombululeni i equation parameter nganye ;

ε - variable random leyo ubonakalise unxaxho phakathi eyona ixabiso theory.