Indlela Yezikwele Ezincane: Ukuhlaziywa Okuphelele
Indlela ye-Least Squares ingenye yezindlela eziyisisekelo kakhulu ekuhlaziyweni kokuhlehliswa kwezibalo, idlala indima ebalulekile ekufakweni kwedatha, ukumodela, kanye nokuhlaziywa kokubikezela. Kuyindlela yezibalo enciphisa isamba sezikwele zomehluko phakathi kwamanani abonwe nabikezelwe. Ngokwenza kanjalo, inikeza umugqa noma ijika elifaneleka kakhulu kusethi ethile yamaphuzu edatha. Kulesi sihloko, sizohlola izisekelo, izinhlelo zokusebenza, okususelwe kukho, kanye nemikhawulo ye-Least Squares Method.
Umlando
Indlela Yezikwele Ezincane yaqala ukwethulwa nguCarl Friedrich Gauss ngo-1795 eneminyaka engu-18, yize ingashicilelwanga kwaze kwaba ngu-1809. U-Adrien-Marie Legendre, esinye isazi sezibalo esidumile, wasungula futhi washicilela indlela efanayo ngokuzimela ngo-1805. Naphezu kwalokhu, ukuqagela kukaGauss nokwandiswa kwendlela kwabeka isisekelo sokusetshenziswa kwanamuhla.
Isisekelo Sezibalo
Ngendlela yayo elula, i-Least Squares Method isetshenziselwa ukuhlanganisa imodeli eqondile nesethi yamaphuzu edatha. Njengoba kunikezwe isethi yamaphuzu edatha angu-`n` \((x_1, y_1), (x_2, y_2), …, (x_n, y_n)\), umgomo ukuthola umugqa \(y = mx + c\) onciphisa isamba sezikwele zamabanga aqondile (ama-residuals) phakathi kwamanani abonwe \(y_i\) kanye namanani abikezelwe umugqa \(\hat{y}_i\).
Ngokwezibalo, le nhloso ingamelwa kanje:
\[ S = \sum_{i=1}^{n} (y_i – \hat{y}_i)^2 = \sum_{i=1}^{n} (y_i – (mx_i + c))^2 \]
Lesi sibalo simelela isamba sezikwele zezinsalela, \(S\), futhi umgomo wethu ukuthola amanani ka-\(m\) (ukwehla) kanye no-\(c\) (ukuvimba) anciphisa u-\(S\).
Ukususwa
Ukuze sithole amanani ka-\(m\) kanye no-\(c\) anciphisa isamba sezikwele \(S\), sithatha ama-derivative angaphelele ka-\(S\) maqondana no-\(m\) kanye no-\(c\) bese siwabeka ku-zero.
1. I-Partial Derivative maqondana ne-\(m\):
\[ \frac{\partial S}{\partial m} = \sum_{i=1}^{n} 2(y_i – (mx_i + c))(-x_i) = 0 \]
2. I-Partial Derivative maqondana ne-\(c\):
\[ \frac{\partial S}{\partial c} = \sum_{i=1}^{n} 2(y_i – (mx_i + c))(-1) = 0 \]
Ukuxazulula lezi zibalo ezihambisanayo kuphumela ezibalweni ezijwayelekile:
\[ m = \frac{n(\sum x_i y_i) – (\sum x_i)(\sum y_i)}{n (\sum x_i^2) – (\sum x_i)^2} \]
\[ c = \frac{(\sum y_i)(\sum x_i^2) – (\sum x_i)(\sum x_i y_i)}{n(\sum x_i^2) – (\sum x_i)^2} \]
Lezi zibalo zisinika amanani ka-\(m\) kanye no-\(c\) anciphisa isamba sezinsalela eziyisikwele.
Izicelo
Indlela ye-Least Squares inezinhlelo zokusebenza eziningi kuzo zonke izizinda ezahlukene:
1. Ezomnotho Nezezimali: Kusetshenziswa ukukhombisa nokubikezela izinkomba zomnotho ezifana ne-GDP, amazinga okwehla kwamandla emali, kanye namanani esitoko.
2. Ubunjiniyela: Kusetshenziswa ekucubungulweni kwesignali, izinhlelo zokulawula, kanye nokuhlolwa kokuthembeka ukuze kufane amamodeli nedatha eqinisekisiwe.
3. Umuthi: Usiza ekuhlanganiseni ama-curve okukhula, ama-curve okuphendula umthamo, kanye nezinye izinqubo zezinto eziphilayo kudatha yokuhlola.
4. Ukufunda Komshini: Amamodeli okubuyela emuva okuqondile, avame ukusebenzisa i-Least Squares Method yokulinganisa amapharamitha, ayisisekelo kumasu okufunda aqondisiwe.
5. I-Astronomy: Isetshenziswa nguGauss ngokomlando ukubala imijikelezo yezinto zasezulwini.
Izinguquko kanye Nezandiso
1. Izikwele Ezincane Ezinesisindo: Ezimweni lapho ukubonwa kunokwehluka okuhlukile, izikwele Ezincane Ezinesisindo zinikeza izisindo endaweni ngayinye yedatha, zinciphisa isamba sezikwele esilinganiselwe.
2. Izikwele Ezincane Ezingezona Eziqondile: Yandisa indlela ukuze ivumelane namamodeli angezona eziqondile, ngokuvamile ngokusebenzisa amasu okuphindaphinda njenge-algorithms ye-Gauss-Newton noma i-Levenberg-Marquardt.
3. Izikwele Ezincane Ezilungisiwe (Ukuhlehla Kwe-Ridge): Kunezela isigwebo esilinganisweni sezikwele ukuvimbela ukulingana ngokweqile, ikakhulukazi okuwusizo lapho kubhekene ne-multicollinearity.
4. Izikwele Ezincane Ezijwayelekile: Zibheka ukubonwa okuhlobene, zandisa indlela yezikwele ezincane kakhulu yokusingatha amaphutha ahlobene ngokuzenzakalelayo kanye nalawo angenawo umkhawulo.
Amathuluzi Okusebenzisa Ikhompyutha
Amathuluzi ahlukahlukene esofthiwe kanye nezindawo zokuhlela zinikeza imisebenzi eyakhelwe ngaphakathi yokwenza i-least squares regression:
1. I-Python: Amalabhulali afana ne-NumPy kanye ne-SciPy ahlinzeka ngemisebenzi ethi `numpy.linalg.lstsq` kanye ne-`scipy.optimize.curve_fit` yokulinganisa izikwele ezincane eziqondile nezingezona eziqondile, ngokulandelana.
2. R: Imisebenzi efana nokuthi `lm()` yamamodeli aqondile kanye no-`nls()` wamamodeli angewona aqondile endaweni yezibalo ephelele ka-R.
3. I-MATLAB: Umsebenzi othi `lsqcurvefit` ku-MATLAB usetshenziselwa ukulingana okungenamugqa, ngokusekelwa okwakhelwe ngaphakathi kwemisebenzi ye-algebra eqondile.
Ukulinganiselwa
Naphezu kokusebenza kwayo okubanzi, i-Least Squares Method inemikhawulo:
1. Ukuzwela Okungaphandle: I-Least Squares ibucayi kakhulu ku-outliers, njengoba igama eliyisikwele likhulisa umthelela wezinsalela ezinkulu.
2. Ukuqagela Ukulingana: Kuqagela ubudlelwano obuqondile phakathi kweziguquguquko, okungenzeka kungabi yiqiniso kumasethi edatha ayinkimbinkimbi.
3. Izinkinga ze-Multicollinearity: Ekubuyeleni emuva okunemigqa eminingi, i-multicollinearity phakathi kwezinguquko zokubikezela ingaphambukisa ukulinganisa kwama-coefficients.
4. Ukufana Kokuhlukahluka (Ukufana): Kucabanga ukuthi ukuhlukahluka kwamagama amaphutha kuhlala njalo kuzo zonke izinto ezibonwayo, okungase kungabi njalo ngaso sonke isikhathi.
Isiphetho
Indlela ye-Least Squares, naphezu kobudala bayo, isalokhu iyitshe elibalulekile ekuhlaziyweni nasekumodeleni idatha yezibalo. Ukulula kwayo, ukusebenza kwayo kahle, kanye nokuzivumelanisa nezimo ezahlukahlukene, kusukela ekuhleleni okuyisisekelo okuqondile kuya kumamodeli ayinkimbinkimbi angewona aqondile najwayelekile, kuyenza ibe usizo kakhulu. Ukuqonda izisekelo zayo zezibalo, ukusetshenziswa kwayo, kanye nemikhawulo kubalulekile kunoma ngubani ohilelekile ekuhlaziyweni kwedatha, ukuqinisekisa ukuthi indlela isetshenziswa ngendlela efanele ukuthola ukuqonda okunengqondo. Kungakhathaliseki ukuthi kusezifundweni noma embonini, indlela ye-Least Squares iyaqhubeka nokuba ithuluzi eliqinile lokuqonda ubudlelwano ngaphakathi kwedatha nokwenza izibikezelo ezinolwazi.
