Nzira yeLeast Squares: Ongororo Yakakwana
Nzira yeLeast Squares ndeimwe yenzira dzakakosha mukuongorora manhamba, ichiita basa rakakosha mukugadzirisa data, kutevedzera, uye kuongorora zvinofanotaura. Inzira yemasvomhu inoderedza huwandu hwemakwere emusiyano uripo pakati pezvakaonekwa nezvakafanotaurwa. Nekuita izvi, inopa mutsetse kana curve yakakodzera kune seti yedata points. Muchinyorwa chino, tichaongorora hwaro, mashandisirwo, zvakatorwa, uye zvipingamupinyi zveNzira yeLeast Squares.
Historical Background
Nzira yeLeast Squares yakatanga kuunzwa naCarl Friedrich Gauss muna 1795 aine makore 18, kunyange zvazvo isina kuburitswa kusvika muna 1809. Adrien-Marie Legendre, mumwe nyanzvi yemasvomhu ane mukurumbira, akagadzira uye akaburitsa nzira yakafanana muna 1805 akazvimirira. Pasinei neizvi, kunyora kwaGauss nekuwedzera nzira iyi kwakaisa hwaro hwekushandiswa kwemazuva ano.
Mathematics Foundation
Muchimiro chayo chiri nyore, nzira yeLeast Squares inoshandiswa kuenderana nemuenzaniso wemutsara kune seti yemapoinzi edata. Zvichipiwa seti yemapoinzi edata e`n` \((x_1, y_1), (x_2, y_2), …, (x_n, y_n)\), chinangwa ndechekuwana mutsetse \(y = mx + c\) unoderedza huwandu hwemapoinzi edaro rakamira (masara) pakati pezvakakosha zvakaonekwa \(y_i\) uye zvakakosha zvakafanotaurwa nemutsetse \(\hat{y}_i\).
Pamasvomhu, chinangwa ichi chinogona kumiririrwa seizvi:
\[ S = \sum_{i=1}^{n} (y_i – \hat{y}_i)^2 = \sum_{i=1}^{n} (y_i – (mx_i + c))^2 \]
Iyi equation inomiririra huwandu hwemasikweya ezvisara, \(S\), uye chinangwa chedu ndechekuwana kukosha kwe \(m\) (slope) uye \(c\) (intercept) kunoderedza \(S\).
Kubviswa
Kuti tiwane kukosha kwe \(m\) uye \(c\) kunoderedza huwandu hwema squares \(S\), tinotora zvikamu zve \(S\) maererano ne \(m\) uye \(c\) tozviisa pa zero.
1. Chikamu Chinobva Pachikamu maererano ne \(m\):
\[ \frac{\partial S}{\partial m} = \sum_{i=1}^{n} 2(y_i – (mx_i + c))(-x_i) = 0 \]
2. Chikamu Chinobva Pachikamu maererano ne \(c\):
\[ \frac{\partial S}{\partial c} = \sum_{i=1}^{n} 2(y_i – (mx_i + c))(-1) = 0 \]
Kugadzirisa ma equation aya panguva imwe chete kunoguma nema equation akajairwa:
\[ m = \frac{n(\sum x_i y_i) – (\sum x_i)(\sum y_i)}{n (\sum x_i^2) – (\sum x_i)^2} \]
\[ c = \frac{(\sum y_i)(\sum x_i^2) – (\sum x_i)(\sum x_i y_i)}{n(\sum x_i^2) – (\sum x_i)^2} \]
Maequation aya anotipa kukosha kwe \(m\) uye \(c\) kunoderedza huwandu hwema residues akaenzana.
Applications
Nzira yeLeast Squares ine mashandisirwo akasiyana-siyana mumadhomini akasiyana-siyana:
1. Zvehupfumi neMari: Zvinoshandiswa kuratidza nekufanotaura zviratidzo zvehupfumi zvakaita seGDP, mitengo yemitengo, uye mitengo yemasheya.
2. Uinjiniya: Inoshandiswa mukugadzirisa zviratidzo, masisitimu ekudzora, uye kuyedza kuvimbika kuti ienderane nedata rekuongorora.
3. Mushonga: Inobatsira mukuenzanisa makomba ekukura, makomba emhinduro yemushonga, uye mamwe maitiro ehupenyu kudata rekuyedza.
4. Kudzidzira Kwemuchina: Mamodheru ekudzokorora mutsara, anowanzo shandisa nzira yeLeast Squares yekufungidzira maparameter, ndiwo musimboti wehunyanzvi hwekudzidza hunotariswa.
5. Astronomy: Yakamboshandiswa naGauss kuverenga kutenderera kwemitumbi yekudenga.
Kusiyana uye Kuwedzera
1. Weighted Least Squares: Kana ongororo dziine musiyano wakasiyana, Weighted Least Squares inopa huremu kune imwe neimwe data point, zvichideredza huwandu hwehuremu hwemasquare.
2. Zvikwere Zvishoma Zvisina Kurongeka: Zvinowedzera nzira yacho kuti ikwane mamodheru asina kurongeka, kazhinji kuburikidza nematekiniki anodzokororwa akadai seGauss-Newton kana Levenberg-Marquardt algorithms.
3. MaSquare Akaderera Akagadziriswa (Ridge Regression): Anowedzera chirango chehuwandu hwemaSquare kudzivirira kukwira zvakanyanya, kunyanya zvinobatsira pakubata nemulticollinearity.
4. MaSquare Akaderera Akajairwa: Anotsanangura ongororo dzakabatana, achiwedzera nzira yeSquare Dzishoma kuti agadzirise zvikanganiso zve autocorrelated uye heteroscedastic.
Zvishandiso zveKomputa
Maturusi akasiyana-siyana esoftware nenzvimbo dzekuronga mapurogiramu anopa mabasa akavakirwa mukati ekuita least squares regression:
1. Python: Maraibhurari akaita seNumPy neSciPy anopa mabasa ekuti `numpy.linalg.lstsq` uye `scipy.optimize.curve_fit` ekugadzirisa zvidimbu zvidiki zviri mumutsara nezvisina mumutsara, zvichiteerana.
2. R: Mabasa akadai se `lm()` emamodheru emutsetse uye `nls()` emamodheru asiri emutsetse munzvimbo yakazara yeR yehuwandu hwehuwandu.
3. MATLAB: Basa rekuti `lsqcurvefit` muMATLAB rinoshandiswa pakugadzirisa zvinhu zvisina kurongeka, nerutsigiro rwakavakirwa mukati rwekushanda kwealgebra yakarongeka.
Nokuremara
Pasinei nekushanda kwayo kwakakura, nzira yeLeast Squares ine miganhu:
1. Kunzwa Zvisina Kunaka: Least Squares inonyanya kunzwa zvisina kunaka, sezvo izwi rekuti squared richiwedzera kukanganiswa kwezvinhu zvikuru.
2. Kufungidzira Kurongeka: Kunofungidzira hukama hwakarongeka pakati pezvimiro, izvo zvingasashanda kune madata akaomarara.
3. Matambudziko eMulticollinearity: Mukudzokorora kwakawanda kwemutsara, multicollinearity pakati pezvinofanotaura zvinogona kukanganisa kufungidzira kwema coefficients.
4. Kuenzana kweKusiyana (Homoscedasticity): Inofungidzira kuti kusiyana kwemashoko ekukanganisa kunogara kuripo pane zvakacherechedzwa, izvo zvingasava zvakadaro nguva dzose.
mhedziso
Nzira yeLeast Squares, pasinei nezera rayo, inoramba iri musimboti mukuongorora nekugadzira data remasvomhu. Kureruka kwayo, kushanda kwayo, uye kugona kwayo kuchinjika kune akasiyana maitiro, kubva pakudzoreredza mutsara kusvika kumamodheru akaomarara asina mutsara uye akajairwa, zvinoita kuti ive yakakosha. Kunzwisisa hwaro hwayo hwemasvomhu, mashandisirwo ayo, uye miganhu yayo kwakakosha kune chero ani zvake ari kuita ongororo yedata, kuve nechokwadi chekuti nzira iyi inoshandiswa nemazvo kuti awane ruzivo rwakakosha. Ingave mudzidzo yepamusoro kana muindasitiri, Nzira yeLeast Squares inoramba iri chishandiso chakasimba chekunzwisisa hukama huri mukati medata uye kuita fungidziro dzakavakirwa paruzivo.
