I-Clustering and Clustering Algorithms: Umhlahlandlela Ophelele, Izinhlobo, Ukusetshenziswa, kanye Nezinzuzo

Informatec Digital » Ama-algorithms » I-Clustering and Clustering Algorithms: Umhlahlandlela Ophelele, Izinhlobo, Ukusetshenziswa, kanye Nezinzuzo

Wake wazibuza ukuthi izinkampani ziyenza kanjani ngokwezifiso imiyalezo yazo kumsebenzisi ngamunye noma ukuthi iNetflix yazi kanjani ukuthi yini okufanele iyincome? Imfihlo isekusetshenzisweni kwama-algorithms okuhlanganisa idatha, inqubo yokuhlaziya idatha esibe yitshe lesisekelo lokufunda komshini kanye nobuhlakani bokwenziwa. Ezweni lanamuhla ledijithali, ukuqonda nokusebenzisa ukuhlanganisa idatha akuvuli nje kuphela umnyango wokuhlukanisa okungcono kodwa futhi kukuvumela ukuthi ubikezele amaphethini, izitayela, kanye nezidingo ezifihliwe ngaphakathi kwedatha.

Kulesi sihloko, uzohlola konke okudingeka ukwazi mayelana nokuhlanganisa: kusukela kulokho okuyikho ngempela nokuthi kusebenza kanjani, kuya kuma-algorithms ahlukene kanye nezicelo zawo ezisebenzayo emikhakheni eyahlukahlukene njengezokwelapha, ukumaketha, i-biology, kanye nokuphepha. Uma usebenza kwisayensi yedatha, ukumaketha, noma ufuna nje ukuqonda ukuthi i-AI iguqula kanjani idatha eluhlaza ibe ulwazi olubalulekile, qhubeka ufunda ngoba lesi yisiqondiso esiphelele nesisesikhathini!

Iyini i-Clustering futhi kungani ibaluleke kangaka?

Ukuhlanganisa, noma ukuhlaziywa kweqembu , kuyindlela yokufunda komshini engagadiwe ekuvumela ukuthi uqoqe izinto, amarekhodi, noma abantu ngokusekelwe ekufaneni kwabo. Umqondo oyinhloko ukuthola amaqembu emvelo ngaphakathi kwesethi yedatha ngaphandle kwamalebula noma izigaba ezichazwe kusengaphambili. Lokhu kudala "amaqoqo" noma amaqembu lapho amalungu afana khona (ngokusho kwezilinganiso zokufana) futhi ahlukile kwabanye.

Le ndlela ibalulekile kumaphrojekthi okufunda komshini ngoba isiza ukuhlola inani elikhulu ledatha, iveze amaphethini afihliwe, inciphise ubunzima, futhi ithuthukise ukwenza izinqumo emabhizinisini. Isetshenziswa esigabeni sokuhlola idatha, ekunciphiseni ubukhulu, ekuhlukaniseni ngaphambi komodeli ogadiwe, noma njengomgomo wokugcina wokwenza ukuhlukaniswa kwemakethe kusebenze kahle kakhulu.

Ezinye izibonelo ezicacile zokuhlanganisa yilezi:

• Khomba izinhlobo zomculo noma iqembu lezingoma ezifanayo ukuze uthole izincomo.
• Hlukanisa amakhasimende ngokususelwe ekuziphatheni kwawo emikhankasweni yokumaketha.
• Ukunciphisa inani lezinto eziguquguqukayo ngokuhlanganisa ubukhulu ekuhlaziyeni kokuhlola.
• Thola okudidayo noma okungaphandle, okufana nokukhwabanisa kwebhange noma ama-spikes angalindelekile kuzinzwa zezimboni.

Okwenza i-clustering ibe ithuluzi elinamandla kangaka ukuthi ayidingi amalebula angaphambilini: i-algorithm ngokwayo ithola isakhiwo sangaphakathi sesethi yedatha, isiza ukubona ukuthi yini ebingenakwenzeka ukuyihlukanisa ekuqaleni.

Ngabe i-clustering isebenza kanjani? Izigaba zenqubo

Inqubo yokuhlanganisa akuyona nje indaba yokusebenzisa i-algorithm nokuyibiza ngosuku: inezigaba eziningana ezenza umehluko phakathi komphumela ophakathi nendawo kanye nokuhlukaniswa okuwusizo ngempela. Ake sibheke izinyathelo ezibalulekile:

1. Ukukhetha nokulungiswa kwedatha: Isinyathelo sokuqala siwukukhetha okuguquguqukayo okufanele kuhlaziywe futhi kuhlanzwe idatha ukuze kuqedwe amaphutha, izimpinda, noma amarekhodi angahambisani. Ikhwalithi yedatha enhle iwukhiye wokuhlanganisa okuthembekile.
2. Ukukhetha i-algorithm (noma inqubo): Kunama-algorithms amaningi, futhi ukukhetha okulungile kuncike ohlotsheni lwedatha, usayizi wayo, ukwakheka kwamaqoqo, kanye nenjongo yokuhlaziya. Yilapho iningi lesayensi engemuva kokuhlangana ilele khona.
3. Incazelo yenani lamaqoqo: Ezinye izindlela zidinga ukuthi ucacise ukuthi mangaki amaqembu azoseshwa, kuyilapho ezinye zinquma lokhu ngokuzenzakalelayo. Lesi sinqumo singenziwa kusetshenziswa imibandela ezenzakalelayo, i-heuristics, noma ngokusekelwe olwazini lwangaphambili lwesizinda.
4. Ukwenza nokuqeqeshwa kwe-algorithm: Ngemva kokusetha amapharamitha, i-algorithm isetshenziswa ukuze kwakhiwe amaqoqo. Ngokuvamile, kwenziwa izivivinyo ezimbalwa, ukulungisa amapharamitha kuze kufinyelelwe iqoqo lekhwalithi.
5. Ukuhlola nokuqinisekisa: Akwanele ukumane uthole amaqoqo; ukubumbana kwazo, ukwehlukana, nokuba wusizo kwazo kufanele kuhlolwe. Amamethrikhi afana nenkomba ye-Silhouette, inertia, kanye nebanga eliphakathi nendawo eliphakathi neqembu liyasetshenziswa.
6. Ukuhunyushwa kwemiphumela nokusebenzisa: Ekugcineni, imiphumela iyatolikwa (yini echaza iqembu ngalinye? Zingasetshenziswa kanjani?) futhi zisetshenziswe ezinhlosweni ezithile ezifana nokuhlukanisa amakhasimende, ukuhlukanisa imikhiqizo, ukwenza imikhankaso ngokugcwele, noma ukwenza izincomo.

Ukuhlanganisa kuyinqubo ephindaphindayo, lapho ukulungisa nokuhumusha kubalulekile ukuze kukhishwe inani langempela kudatha.

Izinhlobo ezahlukene kanye nezindlela zokuhlanganisa

Ama-algorithms okuhlanganisa angahlukaniswa ngezinhlobo eziningana ngokusekelwe ku-logic yawo yangaphakathi kanye nendlela abumba ngayo amaqembu. Ukuqonda kahle lokhu kungafani kuzokuvumela ukuthi ukhethe indlela efanele kakhulu yesimo ngasinye.

• Ukuminyana okusekelwe ekuhlanganiseni: Le ndlela ikhomba amaqoqo njengezifunda zamaphuzu aphezulu, ahlukaniswe izindawo zokuminyana okuphansi. Ivumela ukuthola amaqembu anomumo ongavamile futhi ngokuvamile iziba izinto ezingaphandle noma umsindo. Isibonelo esiyinhloko: I-DBSCAN ne-OPTIC.
• Iqoqo elisuselwa ku-Centroid: Amaphuzu abelwe iqoqo ngokusekelwe ebangeni lawo ukusuka "ku-centroid," emele isikhungo seqoqo. Lokhu ngokuvamile kudinga ukucacisa inombolo yamaqoqo kusengaphambili futhi kuyazwela esikalini sedatha. Izibonelo: K-isho, i-Mini-batch K-isho.
• Iqoqo le-Hierarchical: Yakha isakhiwo esifana nesihlahla (“dendrogram”) ekhombisa ukuthi amaphuzu aqoqana kanjani kancane kancane abe amazinga: kungaba i-agglomerative (phansi phezulu, ukuhlanganisa amaphuzu abe ngamaqembu amakhulu kakhulu) noma ehlukanisayo (kusuka phezulu kuye phansi, ihlukanisa ingqikithi yeqembu ibe amasethi angaphansi).
• Ukuhlanganisa okusekelwe ekusabalaliseni: Isebenzisa amamodeli angenzeka ukuze inqume ubulungu bephoyinti eqenjini ngokubala amathuba okuba liyingxenye yeqoqo ngalinye. Isibonelo sakudala: I-Gaussian Mixture Models (GMM).
• Ukuhlanganisa ngokuhlukanisa: Ihlukanisa idatha kuma-partitions ka-K ukuze iphoyinti ngalinye libe leqembu eliseduze kakhulu ngokuya ngemibandela yebanga. Ama-algorithms anjalo PAM, K-medoids.

Kuye ngohlelo lokusebenza, ivolumu, nokuma kwedatha, uhlobo olulodwa lokuhlanganisa noma olunye luzokhethwa.

Ama-algorithms okuhlanganisa amakhulu nokuthi asebenza kanjani

Ngezansi kunezinye zezindlela ezisetshenziswa kakhulu futhi eziqashelwa kakhulu emikhakheni yokufunda komshini, ukuhlaziywa kwedatha, kanye nobuhlakani bokwenziwa . Ngayinye inezici zayo ezithile, izinzuzo, kanye nemikhawulo:

I-K-Means

I-K-Means iyinkosi yama-algorithms okuhlanganisa ngenxa yobulula bayo kanye nesivinini . Isekelwe ekuchazeni kusengaphambili inani lamaqembu (k) kanye nokwabela iphuzu ngalinye ledatha eqenjini eline-centroid eliseduze kakhulu. Ama-centroid abuyekezwa ngokuphindaphindiwe kuze kube yilapho imisebenzi iyeka ukushintsha.

Izinzuzo: Kulula ukuyisebenzisa futhi ingandiswa. Isetshenziswa kabanzi ekuhlaziyeni okuhlolayo nanjengesethulo sesayensi yedatha.

Okubi: Kudinga ukunquma u-k kusengaphambili, kungahlangana kube yi-optima yendawo, futhi kuyazwela ekuqalisweni kanye nokuma kwamaqembu (kusebenza kabi ngamaqoqo ezimo ezingezona eziyindilinga noma osayizi abahlukene).

I-DBSCAN (Ukuminyana Okususelwe Kwezindawo Zokuhlanganisa izicelo ezinomsindo)

I-DBSCAN ihlonza amaqembu ngokusekelwe ezindaweni eziminyene zamaphuzu futhi iphumelela kakhulu ekutholeni amaqoqo ezimo ezingahleliwe, kanye nokuthola izinto ezingaphandle (umsindo). Akudingi ukucacisa inani lamaqoqo, kodwa kunalokho kudinga amapharamitha amabili: ibanga eliphakeme phakathi kwamaphuzu okufanele abhekwe njengomakhelwane (ama-eps) kanye nenani elincane lamaphuzu ukwakha iqembu.

Izinzuzo: Ithola izimo eziyinkimbinkimbi futhi akudingekile ukuchaza u-k.

Okubi: Isebenza kabi kakhulu kumasethi anobuningi obuguquguqukayo kakhulu futhi idinga ukulungiswa okucophelela kwamapharamitha ukuze kutholakale imiphumela emihle.

Kusho Shift

I-Mean Shift isekelwe "efasiteleni elishelelayo" eliya ezindaweni ezinobuningi bamaphuzu aphezulu, lilungisa ama-centroid aze ahlangane kumamodi (iziqongo zobuningi). Ithola ngokuzenzakalelayo inani lamaqoqo.

Izinzuzo: Akudingi ukucacisa kusengaphambili i-ky futhi kusebenza kahle kudatha yendawo kanye nokubona kwekhompyutha.

Okubi: Ukunciphisa ukukhuliswa kwedatha enkulu kanye nokuncika kusayizi wewindi.

I-algorithm ye-Expectation-Maximization (EM) ene-Gaussian Mixture Models (GMM)

Le algorithm icabanga ukuthi idatha isatshalaliswa ngokwezabelo eziningana zeGaussian, kubalwa amathuba ephuzu ngalinye elingeleqembu ngalinye . Liguquguquka kakhulu kunezindlela ze-K zokuthola amaqembu angewona ayindilinga, futhi iqembu ngalinye lingaba nesimo nosayizi walo.

Izinzuzo: Ifanele izakhiwo eziyinkimbinkimbi kanye nokuhlaziywa okungenzeka.

Okubi: Kudinga ukukhetha inani lezingxenye futhi kungaba nozwela ekuqalisweni.

Omakhelwane be-K-Nearest (KNN) bafake isicelo ku-clustering

Nakuba i-KNN ivame ukusetshenziswa ekuhlukaniseni, ingasetshenziswa futhi ekuhlanganiseni, ukuqoqa amaphuzu ngokwezindawo eziseduze nomakhelwane babo . Kulula, kodwa isikhathi sokubala singaba phezulu njengoba idatha ikhula.

I-Hierarchical Clustering

Ikhiqiza isakhiwo esifana nomuthi (i-dendrogram) esibonisa ukuthi idatha ihlelwe kanjani ngamaqoqo emazingeni ahlukene . Kunezindlela ezimbili eziyinhloko:

• I-Agglomerative (phansi-phezulu): Iphuzu ngalinye ekuqaleni liyiqoqo lalo futhi asondelene kakhulu ahlanganiswa ekuphindaphindweni ngakunye.
• Ukuhlukanisa (phezulu-phansi): Iqala kuqoqo lomhlaba wonke futhi ihlukaniswe ngokulandelana kwama-subset.

Izinzuzo: Awudingi ukucacisa i-ky futhi kuyasiza ekutholeni amazinga angempela kudatha.

Okubi: Inobunzima besikhathi eside futhi ingase ingakhuliswa kakhulu kunezinye izindlela.

I-algorithm ye-BIRCH

I-BIRCH ilungiselelwe amasethi edatha amakhulu kakhulu, anezinombolo . Ifingqa idatha ibe amaqoqo amancane aphakathi lapho kungasetshenziswa khona enye indlela.

Inzuzo eyinhloko: Ukwanda kanye nokuhambisana nezinye izinhlelo zokuhlanganisa.

Ububi: Ayisebenzi kahle ngedatha yezigaba futhi idinga ukucutshungulwa kusengaphambili.

IZINKETHO

I-OPTICS iyisandiso se-DBSCAN esikuvumela ukuthi uthole amaqoqo anobuningi obuhlukene , uhlele amaphuzu ezindaweni eziyinkimbinkimbi zeqembu kangcono.

Ukusakazwa kobudlelwane

Le algorithm ivumela amaphuzu ukuthi "axhumane" ukuze anqume ngezibonelo ezimele futhi akhe amaqembu ngaphandle kokuchaza kusengaphambili ukuthi kuzoba nezingaki . Kuyafaneleka uma singazi ukuthi zingaki izingxenye esifuna ukuzithola.

I-Spectral Clustering

Ngokusekelwe kumbono wegrafu, le ndlela iphatha idatha njengama-node okuthola amaqembu ngokuxhumana kanye nemiphakathi ngaphakathi kwegrafu . Idinga ukubala ama-matrices okufana.

I-algorithm ngayinye inokuhlukahluka kwayo nokuzivumelanisa nezimo, njenge-mini-batch K-means (ishesha idatha enkulu) noma izindlela ze-PAM, i-CLARA ne-FANNY (iwusizo ku-R namasethi amakhulu edatha).

Izinhlelo zokusebenza zangempela zokuhlanganisa kanye nezinzuzo ebhizinisini nobuhlakani bokwenziwa

I-Clustering iguquguquka kangangokuthi ingasetshenziswa kuyo yonke into kusukela kubhayoloji kuya ekukhangiseni okudijithali, ukuphepha, ukunakekelwa kwezempilo, ukuhlelwa kwezinto, kanye nocwaningo:

• Ukuhlukaniswa kwekhasimende: Qoqa abantu ngemikhuba yabo yokuthenga, abakuthandayo, nokuziphatha ukuze wenze imikhiqizo nezinsizakalo zibe ngezomuntu siqu.
• Imithi kanye ne-epidemiology: Kusivumela ukuthi sibone amaphethini ezifweni, siqoqe izithombe ezifanayo zezokwelapha, noma sibikezele izindawo ezinengcuphe ye-epidemiological.
• Ukuhlukaniswa nokuhlelwa kwemikhiqizo: Lungiselela ukuphathwa kwe-warehouse kanye nesakhiwo somkhiqizo ku-e-commerce.
• Ukuqoqwa kwama-athikili nokuqukethwe: Ithuthukisa ukuzulazula nolwazi lomsebenzisi kumawebhusayithi amakhulu nezizindalwazi zesayensi.
• Amanethiwekhi omphakathi nokuhlaziywa komphakathi: Khomba amaqembu abasebenzisi abanentshisekelo efanayo noma amaphethini okusebenzelana.
• Ukukhwabanisa nokutholwa okungaqondakali: Zitholele amaphethini angajwayelekile angakhombisa ukukhwabanisa kwezezimali, amaphutha ezimboni, noma i-cybersecurity.
• Ukuhlukaniswa kwezindawo zejografi: Usizo ocwaningweni lwemakethe lokuhlonza izifunda ezinamandla okuhweba noma ubungozi obuthile.
• I-SEO nokumaketha kokuqukethwe: Amagama angukhiye weqembu nezihloko ukuze uhlonze amathuba futhi udale okuqukethwe okufanele, okuhlosiwe.
• Okuzenzakalelayo kwasekhaya namadivayisi ahlakaniphile: Hlaziya futhi ulungiselele ukusetshenziswa kwensiza ngokuhlanganisa amaphethini okusetshenziswa afanayo.

I-Clustering inikeza ukucaca, kunciphisa ukuzithoba, futhi kusiza ukwenza izinqumo ezingcono ngokusekelwe kudatha yomgomo.

Izinzuzo nezinselele zokusebenzisa ukuhlanganisa ezinkampanini namaphrojekthi wezobuchwepheshe

Izinzuzo eziyinhloko:

• Thuthukisa ukuguqulwa kanye nemikhankaso eqondiwe kangcono: Ngokuhlonza izingxenye ezinembayo, izenzo zokumaketha ziba ngempumelelo kakhulu.
• Khipha ulwazi olufihliwe ebhizinisini: Thola ukufana namaphethini abengeke abonakale ngeso lenyama, okukusiza ukuthi uvule amathuba amasha nezingozi.
• Yehlisa ubungozi: Ukwenza izinqumo ezinolwazi futhi ezihlosiwe kunciphisa amaphutha amasu kanye nokulahlekelwa kwezimali.
• Lungiselela izinqubo nezisetshenziswa: Ngokuhlukanisa idatha nokuthuthukisa iziteshi, unganciphisa izindleko futhi wandise inzuzo.

Izinselelo okufanele zicatshangelwe:

• Isidingo sekhwalithi yedatha enhle: Imiphumela incike kakhulu ekulungiseni nasekuhlanzweni kwedatha yangaphambilini.
• Ukukhetha okufanele kwe-algorithm: Ukungalingani kahle kungaholela emaqenjini angameleli noma angasizi ngalutho.
• Ukuhumusha okulungile: Amaqoqo kufanele enze umqondo webhizinisi futhi angabi amaqoqo angabonakali.
• Ukwehluka: Amanye ama-algorithms awasebenzi kahle ngezigidi zamarekhodi noma izinto zezigaba.

Ukuhlanganisa okuqinile vs. Ukuhlanganisa okuthambile: iyiphi inketho okufanele uyikhethe?

Kuye ngendlela yokwenza, ama-algorithms okuhlanganisa angabela ngokucacile into ngayinye eqenjini elilodwa (iqoqo eliqinile) noma avumele ubulungu obuyingxenye kumaqoqo amaningi (ukuhlanganisa okuthambile noma okungaqondakali).

• Ukuhlanganisa kanzima: Iphuzu ngalinye linikezwe iqoqo ngokuhlukile. Lena indlela enembile kakhulu futhi isetshenziswa izindlela zakudala ezifana ne-K-means.
• Ukuhlanganisa okuthambile: Isici ngasinye sinamathuba okuba semaqenjini amaningana; luwusizo kakhulu ezimeni lapho imingcele phakathi kwamaqembu ingacacile. Isibonelo: Amamodeli engxube ye-Gaussian.

Ukukhetha kuncike enkingeni, idatha, kanye nezinjongo zokuhlaziya.

Izici ezibalulekile zemodeli yokuhlanganisa ephumelelayo

Ukuze ukuhlanganisa kube usizo ngempela, akwanele ukumane usebenzise ama-algorithms ngokungahleliwe. Udinga ukunaka kakhulu:

• Ikhwalithi yedatha nokuhlanzeka: Idatha eyiphutha noma engahambisani ingahlanekezela amaqembu.
• Ukukhetha okuguquguqukayo: Ukukhetha ubukhulu obufanele kubalulekile ukuze uthole amaqoqo abamele.
• Chaza kahle inani lamaqembu: Uma inombolo engalungile ikhethiwe, amaqembu angase angasebenzi.
• Qinisekisa imiphumela: Sebenzisa amamethrikhi afanelekile futhi, uma kungenzeka, ochwepheshe bebhizinisi ukuze uqinisekise incazelo yamaqembu.
• Phinda futhi ulungise: Ukuhlanganisa akuvamisile ukucacisa okokuqala: imizamo embalwa ngokuvamile iyadingeka ukuze ulungise imodeli.

Ukuhlanganisa ukumaketha kokuqukethwe kanye ne-SEO: Thola amathuba amasha

Ukuhlanganisa akusizi nje ukuqoqa amakhasimende noma imikhiqizo; ingaphinda iguqule okuqukethwe kwakho kanye namasu we-SEO:

• Thola izihloko ezifanele: Ngokuhlanganisa amagama angukhiye nezihloko, ungakwazi ukuhlonza amaphethini okusesha kanye nezitayela ozithakaselayo.
• Lungiselela ukwakheka kokuqukethwe: Kuyasiza ukudala ama-silos anezihloko nokuthuthukisa ukuxhumanisa kwangaphakathi, ukwandisa isikhathi ekhasini kanye negunya lewebhusayithi.
• Gxilisa isu lakho legama elingukhiye: Ikuvumela ukuthi uthuthukise amaqoqo egama elingukhiye futhi udale amakhasi athile okufika eqembu ngalinye, uthuthukise ukuma.
• Izithameli zengxenye: Ngokuhlaziya amaphethini okuziphatha, okuqukethwe kungadalwa ngendlela efanele amaphrofayili abasebenzisi ahlukene.

Ukuhlanganisa kwenza okuqukethwe kuhambisane kakhulu, kwenziwe kube ngokwakho, futhi kusebenze kahle, kokubili kumsebenzisi naku-algorithm ye-Google.

Yimaphi ama-algorithms akhona futhi ungayikhetha kanjani efaneleka kakhulu?

Ukukhetha i-algorithm ye-clustering kuncike ku:

• Usayizi nemvelo yedatha (izinombolo, ngokwezigaba, indawo, njll.).
• Umumo olindelekile wamaqoqo (i-spherical, i-arbitrary, hierarchical, njll.).
• Ukuba khona komsindo noma izinto ezingaphandle.
• I-scalability nesivinini esidingekayo ukuze kuhlaziywe.

Nakuba i-K-means ilungele amasethi edatha amakhulu ezinombolo kanye namaqoqo ayindilinga, i-DBSCAN kanye ne -OPTICS ziyaphumelela ngezimo eziyinkimbinkimbi nomsindo. Ukuhlanganiswa kwezigaba akunakuqhathaniswa uma sidinga ukuqonda isakhiwo sobudlelwano phakathi kwamaqoqo, futhi kuwusizo ikakhulukazi ezimweni zokungaqiniseki.

Ngezinye izikhathi kuyasiza ukuhlanganisa izindlela eziningana: isibonelo, ukusebenzisa amasu anjenge-BIRCH noma i-Mini-batch K-kusho ukunciphisa ivolumu yedatha bese usebenzisa i-algorithm ethuthukisiwe kumaqoqo aphumayo.

Ukuqaliswa okungokoqobo: izibonelo nekhodi ku-Python

Ngokuthambekela kobuchwepheshe ngokwengeziwe, ngezansi sabelana ngamazwibela enziwe lula (ku-Python nokusebenzisa i-Scikit-learn) kwamanye ama-algorithms okuxoxwe ngawo. Ngale ndlela, ungazizwela wena ukuthi ukuhlanganisa kusebenza kanjani ekusebenzeni.

I-K-Means

from sklearn.cluster import KMeans
model = KMeans(n_clusters=3)
resultados = model.fit_predict(datos)

I-DBSCAN

from sklearn.cluster import DBSCAN
modelo = DBSCAN(eps=0.5, min_samples=5)
resultados = modelo.fit_predict(datos)

I-Hierarchical Clustering

from sklearn.cluster import AgglomerativeClustering
modelo = AgglomerativeClustering(n_clusters=3)
resultados = modelo.fit_predict(datos)

Amamodeli Engxubevange ye-Gaussian

from sklearn.mixture import GaussianMixture
modelo = GaussianMixture(n_components=3)
modelo.fit(datos)
resultados = modelo.predict(datos)

Kusho Shift

from sklearn.cluster import MeanShift
modelo = MeanShift()
resultados = modelo.fit_predict(datos)

Ungalungisa amapharamitha njengenombolo yamaqembu, ibanga, iwindi, njll., kuye ngedathasethi yakho nezinjongo zakho.

Amathiphu abalulekile namaphutha okufanele uwagweme ekuhlanganiseni

• Ungenzi ngokwejwayelekile noma ukale idatha: Kubalulekile ukuthi amabanga aqhathaniswe nokuthi ukuhlanganisa kusebenze.
• Ukulinganisa ngokweqile umthamo we-algorithm: Ayikho indlela engenaphutha, futhi ukutolika kweqoqo kufanele ngaso sonke isikhathi kwenziwe ngomqondo webhizinisi.
• Ziba ukuqinisekiswa: Amaqoqo kufanele ahlolwe ngobuningi nangekhwalithi ngaphambi kokwenza izinqumo zamasu ngokusekelwe kuwo.
• Ukucabanga ukuthi kunomphumela owodwa kuphela osebenzayo: Ukuhlanganisa kuvame ukuhlola; izigaba eziningana kungase kube nengqondo, kuye ngokuthi inhloso.

Okubalulekile wukuphindaphinda, ukuhlaziya, nokuqonda kokubili ngokobuchwepheshe nangokuhlakanipha kwebhizinisi.

Ngokuhlanganisa, izinkampani nezingcweti zanoma yimuphi umkhakha zingasebenzisa inani elifihliwe kudatha yazo, zithole amaphethini angalindelekile, futhi zithuthukise kokubili amasu nemiphumela yazo. Kusukela ekuhlukaniseni okulungiswe kahle kuya ekuthuthukiseni izinqubo zangaphakathi noma ukuhlola amathuba amasha emakethe, ukuhlanganisa ama-algorithms asephenduke insika yezibalo zesimanje.