Désequilibre ci klaas yi ak jëlaat échantillonnage
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.
Résumé
Resampling rebalances the training data so the model actually learns to spot the minority.
Plongeur bu xóot
Sudee klaas yi dañu jaxasoo, ab model mën na yegg ci 99.9% ci njubte ci saa yu nekk muy wax lu ëpp te du musa jàpp benn njuuj njaaj, te loolu amul benn njariñ. Resampling dafay defar seddaleb tàggat yaram ci ñaari anam yu yaatu. Oversampling dafay ñaari yoon wala dafay boole misaal yu néew - SMOTE (Synthetic Minority Over-sampling Technique) dafay sos poñ yu bees ci boole misaal bu néew ak dëkkandoo yu néew yi gëna jege, moo gën ñu ko kopie. Undersampling lu moy loolu dafay sànni misaal yu bari (ci anam wu mujjee, wala ci anam wu xarañ jaaraleko ci lëkkalekaay Tomek wala NearMiss) ba ci mbir yi, ci njëgu sànni done. Alternatif yi nga moytu laal done yi bokkuna ci pondération classe (penaliser njuumte yu ndaw yi gëna bari ci fonction perte) ak yamale dogal threshold ginaaw tàggat.
Gis-gis xarala
Regle bu am solo: jëlaat misaalu ensemble tàggat bi kese, bul musa jëfandikoo ensemble validation wala test, te saa yu nekk jëlaat misaal ci biir pli cross-validation. Oversampling balaa ñuy xaaj leaks yu jege-ñaari poñ ci biir test bi ba noppi di yokk poñ yi. Ndax njub amul benn njariñ fii, jàngat bi dafa wara sukkandikoo ci njub, fàttaliku, F1, AUC bu njub, wala Matthews Correlation Coefficient - metrics yuy des njub sudee klaas bu baax bi bariwul.
njeextalu pexe
Njëgg ak budget
Dogal yi architecture di jël dañuy indi njariñ ak njëgu liggéey bi ay at ci ginaaw.
dogal yu gëna leer
Njàngalem xarala yi dafay jàppale ekip yi ñu tànn li gën, te baña yam ci li gëna bees daal.
Xool kalite
Tanneef yu gëna baax ci wàllu ingeñër dina wàññi jafe-jafe yi ci wàllu wóor ci liggéey bi.
Ëlëgu désequilibre ci klaas yi ak jëlaat misaal yi
Resampling mingi gëna otomatise ci biir pipeline ML, ak bibliotek yu melni imbalanced-learn di boole ci njubluwaay buñ jaar. Gëstu dafay toxu ci jàng bu xarañ ci njëg ak liggéey yuñ defar ngir ñàkk - lu melni ñàkk focal, luy wàññi misaal yu yomb yu bari - luy faral di gëna am njariñ ci resampling bu xóot ci reso yu xóot yi. Ngir done tablo ak nataal, xeetu generatif yuy boole misaal yu néew yu dëggu ñu ngi feeñ nekk wuutu bu gëna xarañ ci interpolation bu nuroo ak SMOTE.
Doxal ci àdduna dëgg
Taggat ab detektëru njuuj njaaj ci kàrtu kredi fu njuuj njaaj dëgg nekk ci suufu 1% ci jëflante yi, jëfandikoo SMOTE ngir yokk njuuj njaaj yu bariwul
Tabax ab xeetu pajum feebar bu bariwul lumuy am ci malaad yu néew ci teemeer boo jél, jëfandikoo ay poñ ci klaas suko defee ñu daan bu baax ñi ñu ñakk
Gis mbir yu defoon ay jafe-jafe ci ligne de fabrication bi daanaka produit yépp di jaar ci inspection, di wàññi mbir yu 'baax yi' ngir yemale tàggat yaram
Dañuy màndargaal ay njuumte yu bari ci reso bi ci biir këyitu kaaraange siber, te trafik bu jaar yoon moo ko ëpp doole, ñu jàngat ko ci Precision-Recall AUC ci barabu njub
Risk yi ak balustrade yi
Optimize benn benchmark mën na nëbb ñakk kattan yu gëna yaatu ci sistem bi.
Njëg li ñuy fay ci infrastructure yi ak ci toppatoo dañuy faral di suufeel.
Bu sistem yi di gëna xawa jafee xam, jafe-jafe yi am ci wàllu kaaraange ak seetlu mën nañu gëna bari.
Roadmap ngir samp gi
Mandargal latency, kalite, ak njëg yi laata ngay jëfandikoo.
Benchmark ci biir sargal ak done yu dëggu.
Jumtukaay bi di saytu njuumte yi, derive bi ak njeextalu jëfandikukat bi.
Waajal rollback ak yooni tontu ci jafe-jafe yi laata ngay eskale.
Weyal di banneexu
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
Gis bi ci topp
Reseau Siamese ak ñàkka am ñatti doom
Laaj yi ñuy faral di laaj
What is Class Imbalance and Resampling?
Désequilibre ci klaas mooy su benn njariñ ëppee beneen - lu melni 99.9% jëflante yu baax ak 0.1% njuuj njaaj - luy nax model yi ngir ñu baña bàyyi xel ci klaas bu bari waaye am solo. Resampling dafay ekilibrewaat done yiñ tàggat suko defee model bi jàng gis ñu néew ñi.
Lan moo waral njubte bu leer nekkul metrik bu baaxul ci jafe-jafe xaaj bu jaxasoo lool?
Sudee 99.9% ci jafe-jafe yi dañu negatif, model biy faral di wax lu baaxul dafay am 99.9% ci njubte ci jàpp zero positive, kon njubte gi dafay nëbb ñàkka mëna dem ci klaas bu bariwul.
Luy SMOTE?
SMOTE (Pexem Synthetic Minority Over-Sampling) dafay sos misaali minorité yu bees ci interpolation ci digganté benn point minorité ak dëkkandoo yu ndaw yi gëna jege, moo gën ñu leen di ñaari yoon.
Ban gis-gis mooy saafara jafe-jafe te du soppi limu misaali tàggat yi?
Pondération classe dafay bàyyi done yi ñu laal leen, lu moy loolu dafay tax fonction perte bi gëna yar njuumte yi ci classe minorité bi.
Lan mooy risk bu mag bi ci jëfandikoo oversampling balaa ngay xaaj say done ci saxaar ak ensemble test?
Resampling balaa xaaj bi dafay may poñ yu néew yu nuru yu nuru ñu feeñ ci saxaar gi ak ci test yi, di sacc xibaar yi ba noppi di génne ay performance yu ëpp yaakaar, yu amul dëgg.
Lan mooy jafe-jafe bi gëna mag ci sampling bu bariwul?
Undersampling dafay ekilibre ay klaas ci sànni misaal yu bari, te loolu mën na sànni done yu am njariñ ba noppi gàllankoor kàttanu model bi ci jàng klaas yu bari.