Amasu Okubona Idatha Kuzibalo
Ukubona idatha kuyisici esibalulekile sezibalo zanamuhla. Phakathi kobuningi bedatha evela ezinhlolovo, ekuhlolweni, ekuthengisweni kwedijithali, nasezinzwa, inselelo eyinhloko akukhona nje ukuqoqa izinombolo, kodwa futhi nokuveza incazelo ngemuva kwazo. Yilapho ukubonakala kwedatha kungena khona: ukuguqula idatha eluhlaza ibe yizibonisi eziqondakalayo, ukusiza ukuthola amaphethini, ukuthola okungafani, ukuqhathanisa amaqembu, nokusekela ukwenza izinqumo ezisekelwe ebufakazini. Lesi sihloko sixoxa ngamasu avamile okubona idatha ezibalweni, izinhloso zawo, kanye nezimiso zokuqinisekisa ukuthi ukubonakala akudukisi.
1. Indima Yokubona Idatha Kuzibalo
Kuzibalo, ukubona ngeso lengqondo kusetshenziswa ezigabeni eziningana zokuhlaziya. Okokuqala, ngesikhathi sokuhlaziywa kwedatha yokuhlola (i-EDA), amagrafu asiza abahlaziyi ukuqonda ukusatshalaliswa, ukuhlukahluka, izinto ezingaphandle, kanye nobudlelwano phakathi kweziguquguquko ngaphambi kokwakha imodeli. Okwesibili, ukubona ngeso lengqondo kusetshenziselwa ukudlulisa imiphumela: imibiko yocwaningo, izethulo zebhizinisi, kanye nezincwadi zesayensi zidinga amagrafu acacile ukuze kuqondwe ngokushesha okutholakele okubalulekile. Okwesithathu, ukubona ngeso lengqondo kubalulekile nasekuqinisekisweni kwemodeli, isibonelo, ngokusebenzisa amagrafu asele, amagrafu abikezelwe uma kuqhathaniswa nawangempela, noma amanye amagrafu okuxilonga.
Ngamanye amazwi, ukubona idatha akuyona nje "umhlobiso" wemibiko, kodwa kuyithuluzi lokuhlaziya elithonya ikhwalithi yeziphetho zezibalo.
2. Amasu Okubona Ngeso Ledatha Engashintshi (I-Variable Eyodwa)
Uma ukugxila kokuhlaziywa kungokwesiguquli esisodwa kuphela, umgomo oyinhloko ukuqonda isimo sokusabalalisa, isikhungo sedatha, kanye nokusabalala kwaso.
a. I-Histogram
I-histogram ibonisa ukusatshalaliswa kwedatha yezinombolo ngokuhlukanisa idatha ngezikhawu (ama-bin). Kusukela ku-histogram, singanquma ukuthi idatha iyalingana, igobekile ngakwesokudla noma kwesobunxele, noma ineziqongo ezingaphezu kweyodwa (ezine-multimodal). Ukukhetha inani lama-bin kubalulekile: ama-bin ambalwa kakhulu angafihla imininingwane, kuyilapho ama-bin amaningi kakhulu angenza igrafu ibe nomsindo.
b. Isakhiwo sebhokisi (umdwebo webhokisi)
I-boxplot ifingqa idatha isebenzisa i-median, i-quartiles, kanye ne-outliers. Le ndlela iwusizo kakhulu ekuboneni ngokushesha ukuhlukahluka kanye ne-outliers. Kuzibalo, ama-boxplot ayithuluzi elijwayelekile lokuqhathanisa ukusatshalaliswa phakathi kwamaqembu ngenxa yobuncane bawo.
c. Isakhiwo sobuningi (ijika lobuningi)
Ijika lobuningi lifana ne-histogram kodwa libushelelezi. Lokhu kubuka kusiza ukukhombisa isimo sokusabalalisa ngaphandle kokuthembela kakhulu ekukhetheni ibhini. Ama-density plots avame ukusetshenziswa ukuqhathanisa ukusabalalisa okubili kugrafu eyodwa.
d. Ishadi lebha ledatha yezigaba
Ngeziguquguquko zesigaba (isb., ubulili, izinga lemfundo, isigaba somkhiqizo), amashadi ebha ayisinqumo esikhethwayo. Qiniseka ukuthi ama-axes esigaba ahlelwe ngokunengqondo (isb., ngokulandelana kwemvelo noma imvamisa) ukuze kube lula ukuwafunda.
3. Amasu Okubona Ngokubona Ukuqhathanisa Phakathi Kwamaqembu
Ezifundweni eziningi zezibalo, sidinga ukuqhathanisa idatha ngamaqembu (isib. ukwelashwa vs ukulawula, isifunda A vs B, noma amakilasi amaningi).
a. Ibhokisi lebhokisi ngeqembu ngalinye
Ama-boxplot asebenza kahle kakhulu ekuqhathaniseni ama-median, ukusatshalaliswa, kanye nama-outliers phakathi kwamaqembu. Uma kunamaqembu amaningi, cabanga ngokujikelezisa ama-axes noma ukusebenzisa ukuma okuvundlile ukuze wenze amalebula afundeke kalula.
b. Isakhiwo se-violin
Isakhiwo se-violin sihlanganisa i-boxplot kanye ne-density curve ukuze kuboniswe kokubili isifinyezo kanye nesimo sokusabalalisa. Lokhu kuyasiza uma sifuna ukunquma ukuthi umehluko weqembu ubangelwa yizimo ezahlukene zokusabalalisa.
c. Isakhiwo sesilinganiso/sephuzu esinezibha zamaphutha
Ukuze uthole ukuxhumana okunengqondo okwengeziwe, ukuhlela isilinganiso ngamabha wamaphutha (isb., ukuphambuka okujwayelekile, iphutha elijwayelekile, noma isikhawu sokuzethemba) kugcizelela isilinganiso esilinganiselwe kanye nokungaqiniseki kwaso. Kodwa-ke, le ndlela idinga ukuqaphela: isilinganiso singafihla ukusatshalaliswa okungalingani noma okune-multimodal.
4. Amasu Okubona Ngombono Wobudlelwano Phakathi Kweziguquguquko Ezimbili (I-Bivariate)
Ukuhlaziywa kwe-Bivariate kuhlose ukuqonda ubudlelwano phakathi kweziguquguquko ezimbili, kungaba ngezinombolo-zezinombolo, ngezinombolo-zezigaba, noma ngezigaba-zezigaba.
a. Isakhiwo se-scatter (umdwebo we-scatter)
I-scatter plot iyindlela evame kakhulu yokukhetha iziguquguquki ezimbili zezinombolo. Ibonisa amaphethini aqondile, angaqondile, ahlanganisiwe, kanye nama-outliers. Ekuhlaziyweni okuthuthukisiwe, ama-scatter plot avame ukuhambisana nemigqa yokubuyela emuva noma ukushelela (isb., i-LOESS) ukuze kucaciswe izitayela.
b. Ishadi lomugqa ledatha yochungechunge lwesikhathi
Uma i-variable yezinombolo ishintsha ngokuhamba kwesikhathi, igrafu yomugqa isiza ukuhlonza izitayela, ubunjalo besizini, kanye nokuphakama. Ezibalweni zochungechunge lwesikhathi, lesi sakhiwo sivame ukusetshenziswa ngaphambi kokwenza imodeli ye-ARIMA, ukushelela kwe-exponential, noma amanye amamodeli.
c. Imephu yokushisa yezigaba noma ama-matrices
Ama-heatmap afanelekela amathebula e-contingency noma ama-correlation matrices. Imibala imelela ubukhali noma ubukhulu bamanani. Le ndlela iyasebenza kumasethi amakhulu edatha, kodwa uhlelo lombala kumele lukhethwe ngendlela efanele ukuze kugwenywe ukuchazwa kabi.
5. Amasu Okubona Izinto Eziningi (Iziguquguquko Ezingaphezu Kwezimbili)
Uma idatha ihilela iziguquguquko eziningi, inselele iwukubonisa ulwazi ngaphandle kokwenza igrafu ibe yinkimbinkimbi kakhulu.
a. Isakhiwo sokuhlukanisa ngombala/usayizi/isimo
I-scatterplot ingabonisa iziguquguquko ezengeziwe kusetshenziswa umbala (owesigaba), usayizi wamachashazi (owezinombolo), noma ukuma (owesigaba). Le ndlela inamandla, kodwa idinga ukuqinisekisa ukuthi inganekwane icacile nokuthi umehluko obonakalayo awucashile kakhulu.
b. Ukuhlanganisa ubuso (iziphindaphindo ezincane)
Ukuhlukanisa ishadi elifanayo libe amaphaneli amaningi ngokusekelwe ezigabeni (isb., ngesifunda noma ngohlobo lomkhiqizo). Lokhu kuvame ukusebenza kangcono kunokufaka ulwazi oluningi kakhulu eshadini elilodwa.
c. Isakhiwo sebhangqwana/isakhiwo se-scatterplot matrix
I-scatterplot matrix ibonisa zonke izibhangqwana zeziguquguquko zezinombolo kugridi eyodwa. Lokhu kusiza ukuhlonza ubudlelwano phakathi kweziguquguquko, ubudlelwano obuqinile, noma amaphethini eqoqo. Le ndlela iwusizo ku-EDA, ikakhulukazi ngaphambi kokuhlaziywa kokuhlehla noma ukufunda komshini.
d. Iploti ye-PCA biplot noma iploti yokunciphisa ubukhulu
Kumasethi edatha anezici eziningi, amasu okunciphisa ubukhulu njenge-PCA, i-t-SNE, noma i-UMAP angahlanganisa idatha ku-2D ukuze kwembulwe amaqoqo. Kuzibalo, ama-biplot e-PCA angabonisa futhi umnikelo weziguquguquko ezingxenyeni eziyinhloko. Kodwa-ke, imiphumela yokunciphisa ubukhulu kufanele ihunyushwe ngokucophelela ngenxa yokulahleka kolwazi ngesikhathi sokucindezelwa.
6. Izimiso Zokuklama Ukubona Okuhle
Indlela enhle yokubona izinto ngeso lengqondo ingaba yingozi uma umklamo ungalungile. Nazi ezinye izimiso ezibalulekile kumongo wezibalo:
1. Khetha ishadi elifanela uhlobo lwedatha yakho kanye nombuzo wakho. Ungasebenzisi ishadi lephayi ezigabeni eziningi kakhulu noma ukuqhathanisa umehluko omncane.
2. Sebenzisa izikali ze-axis ngokwethembeka. Ukunquma ama-axes (isb., i-y-axis ayiqali ku-zero) kungakhulisa umehluko ngokubonakalayo. Lokhu ngezinye izikhathi kuyamukeleka, kodwa kufanele kunikezwe umongo kanye nesizathu.
3. Naka umbala. Sebenzisa i-palette evumelana nombala futhi ugweme ukuhlanganiswa kwemibala okunzima.
4. Nikeza amalebula nemithombo ecacile. Isihloko, inganekwane, amayunithi, kanye namagama-ncazo kumele kuphelele ukuze igrafu ikwazi ukuzimela yodwa.
5. Bonisa ukungaqiniseki uma kufanele. Ezibalweni zokuqagela, ukubonisa izikhawu zokuzethemba, amabhendi okubikezela, noma imigoqo yamaphutha kunolwazi oluningi kunokubonisa izinombolo ezilodwa.
6. Gwema "izinto ezingadingekile zeshadi." Imiphumela ye-3D, ukuhlotshiswa ngokweqile, noma ama-gradient angadingekile kungaphazamisa umyalezo oyinhloko.
7. Amaphutha Avamile Ekubonweni Kwezibalo
Amanye amaphutha avamile futhi angadukisa abafundi:
– Ukusebenzisa isilinganiso sedatha egobile kakhulu ngaphandle kokubonisa i-median noma ukusatshalaliswa.
– Ukuhlanganisa izigaba eziningi kakhulu kwenza igrafu ibe nzima ukuyifunda.
– Ayibonisi usayizi wesampula yize ukuqhathaniswa kweqembu kungaba nobandlululo uma inani ledatha lihlukile kakhulu.
– Ukuthola iziphetho eziyimbangela ezivela ezindabeni ezihlakazayo ezibonisa ukuhlobana kuphela.
8. Isiphetho
Amasu okubona idatha kuzibalo ahlanganisa ukuqonda idatha, izinhloso zokuhlaziya, kanye namakhono okuxhumana. Ama-histogram, ama-boxplot, ama-scatterplot, amashadi omugqa, ama-heatmap, ngisho namasu e-multivariate afana nokuhlukanisa kanye ne-PCA asiza abahlaziyi ukuthi babone okungabonakali kalula ematafuleni ezinombolo. Kodwa-ke, ukubona okuhle akugcini nje ngokuba "kuhle" kodwa futhi kunembile, kuthembekile, futhi kugxile emibuzweni abafuna ukuyiphendula.
Ngokusebenzisa amasu afanele kanye nezimiso zokuklama okuzwakalayo, ukubona idatha kungaba yibhuloho elinamandla phakathi kokuhlaziywa kwezibalo okuyinkimbinkimbi nokuqonda okulula kwezilaleli ezahlukahlukene—kusukela kubacwaningi kuya kubenzi bezinqubomgomo.
Uma ungathanda, ngingaguqula lesi sihloko sibe sezemfundo kakhulu (ngezinkomba), ngengeze izibonelo zamacala, noma ngenze inguqulo egxile kwisofthiwe ethile efana ne-Excel, R, noma i-Python.
