Kuringaniza ibyiciro no gusubiramo
Class imbalance is when one outcome vastly outnumbers another — like 99.9% legitimate transactions versus 0.1% fraud — which tricks models into ignoring the rare but important class.
Incamake
Resampling rebalances the training data so the model actually learns to spot the minority.
Kwibira cyane
Iyo amasomo agoramye, icyitegererezo gishobora gukubita 99.9% muburyo bwo guhanura ubwinshi kandi ntibigere bifata uburiganya bumwe, ntacyo bumaze. Gusubiramo bikosora amahugurwa yo kugabana muburyo bubiri bwagutse. Kurengaho kwigana cyangwa guhuza ingero nkeya - icyiciro cya SMOTE (Synthetic Minority Over-sampling Technique) gitanga ingingo nshya muguhuza icyitegererezo cyabato nabaturanyi bacyo bakegereye aho kubigana. Kudasiba ahubwo birukana ingero nyinshi (kubushake, cyangwa mubwenge ukoresheje uburyo nka Tomek ihuza cyangwa NearMiss) kugeza kubintu hanze, kubiciro byo guta amakuru. Ubundi buryo bwirinda gukoraho amakuru burimo uburemere bwibyiciro (guhana amakosa ya bake mumikorere yigihombo) no guhindura imipaka nyuma y amahugurwa.
Ubushishozi
Itegeko rikomeye: ntangarugero gusa amahugurwa yashizweho, ntanarimwe yemeza cyangwa igeragezwa, kandi burigihe ntangarugero imbere yambukiranya imipaka. Kurengana mbere yo gutandukanya ibimeneka hafi-kwigana amanota mugupimisha no kuzamura amanota. Kuberako ubunyangamugayo ntacyo busobanuye hano, isuzuma rigomba gushingira kubisobanuro, kwibutsa, F1, Precision-Kwibuka AUC, cyangwa Coefficient ya Matayo Matayo - ibipimo bikomeza kuba inyangamugayo mugihe icyiciro cyiza kidasanzwe.
Ingaruka z'Ingamba
Igiciro na bije
Ibyemezo byubwubatsi bitwara imikorere nigiciro cyimikorere kumyaka.
Ibyemezo bisobanutse
Ubuhanga bwa tekinike bufasha amakipe guhitamo umurongo ukwiye, ntabwo ari shyashya gusa.
Kugenzura ubuziranenge
Guhitamo neza bya injeniyeri bigabanya ibintu byizewe mubikorwa.
Kazoza Kuringaniza Kuringaniza no Gusubiramo
Gusubiramo bigenda byikora imbere mumiyoboro ya ML, hamwe namasomero nkuburinganire-wige kwinjiza muburyo butandukanye. Ubushakashatsi bugenda bugana ku myigire-yimyigishirize yimyigishirize hamwe nigikorwa cyo gutakaza igihombo - nko gutakaza intumbero, ibyo bikaba byerekana uburemere bworoshye ingero nyinshi - akenshi biruta ibicuruzwa biva mu miyoboro yimbitse. Kubisobanuro nimbonerahamwe yamakuru, moderi yibyara igereranya ingero zifatika zifatika zigaragara nkumusimbuzi ukomeye cyane wa SMOTE-interpolation.
Gushyira mu bikorwa Isi
Guhugura ikarita yinguzanyo-ikarita yerekana uburiganya aho uburiganya nyabwo buri munsi ya 1% yubucuruzi, ukoresheje SMOTE kugirango wongere imanza zidasanzwe.
Kubaka icyitegererezo cyubuvuzi ku ndwara idasanzwe igaragara ku barwayi bake ku ijana gusa, gukoresha uburemere bw’ibyiciro bityo ibibazo byabuze bihanwa cyane
Gutahura ibintu bifite inenge kumurongo wo gukora aho ibicuruzwa hafi ya byose byatsinze igenzura, ugasobanura ibintu 'byiza' kugirango uhuze imyitozo
Kumenyekanisha imiyoboro idasanzwe yibikoresho byumutekano wibikoresho byiganjemo traffic isanzwe, bisuzumwa na Precision-Ibuka AUC aho kuba ukuri
Ingaruka & Kurinda
Gutezimbere igipimo kimwe gishobora guhisha intege nke za sisitemu.
Ibikorwa Remezo no kubungabunga akenshi usanga bidahabwa agaciro.
Icyuho cyumutekano no kwitegereza birashobora kwiyongera uko sisitemu igenda igorana.
Igishushanyo mbonera
Sobanura ubukererwe, ubuziranenge, nigiciro cyibiciro mbere yo kubishyira mubikorwa.
Ibipimo byerekana umutwaro ufatika hamwe namakuru yimiterere.
Gukurikirana ibikoresho kubikosa, drift, ningaruka zabakoresha.
Tegura inzira yo gusubiza ibyabaye mbere yo gupima.
Komeza Ubushakashatsi
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Class Imbalance and Resampling quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Ubuyobozi bukurikira
Imiyoboro ya Siamese hamwe nigihombo cya gatatu
Ibibazo bikunze kubazwa
What is Class Imbalance and Resampling?
Ubusumbane mu byiciro ni mugihe igisubizo kimwe cyarushije cyane ikindi - nka 99.9% byemewe byemewe na 0.1% uburiganya - buriganya icyitegererezo cyo kwirengagiza icyiciro kidasanzwe ariko gikomeye. Kongera gusubiramo amakuru yimyitozo kuburyo icyitegererezo cyiga kumenya umubare muto.
Ni ukubera iki ubunyangamugayo busobanutse neza kubipimo bidahwitse cyane?
Niba 99,9% yimanza ari mbi, icyitegererezo gihora gihanura ibibi kibona 99.9% byukuri mugihe ufata ibyiza bya zeru, kubwibyo rero guhisha kunanirwa rwose kurwego rudasanzwe.
SMOTE ikora iki?
SMOTE (Synthetic Minority Over-sampling Technique) itanga ingero nshya zabantu bake muguhuza ingingo ya bake nabaturanyi bayo bakegeranye, aho kuyigana gusa.
Nubuhe buryo bukemura ubusumbane NTA guhindura umubare wingero zamahugurwa?
Uburemere bwibyiciro busiga amakuru adakorwaho ahubwo bigatuma imikorere yigihombo ihana amakosa kumatsinda mato cyane.
Ni izihe ngaruka zingenzi zo gukoresha ibicuruzwa mbere yo kugabanya amakuru yawe muri gari ya moshi no mu bizamini?
Ongera usubiremo mbere yo gutandukana ureke hafi-ya-amanota mato mato agaragara haba muri gari ya moshi no mu bizamini, gutangaza amakuru no gutanga umusaruro ushimishije, udashyitse.
Ni izihe ngaruka nyamukuru zo kudasobanura ibintu?
Kudashyira mu gaciro kuringaniza amasomo mu guta ingero nyinshi, zishobora guta amakuru yamakuru kandi bikangiza ubushobozi bwicyitegererezo cyo kwiga ibyiciro byinshi.