Pregled bibliografske jedinice broj: 926527
Cost-sensitive learning from imbalanced retail credit dataset
Cost-sensitive learning from imbalanced retail credit dataset // International Scientific Conference on IT, Tourism, Economics, Management and Agriculture – ITEMA 2017 / Nedanovski, Pece ; Filipović, Dejan ; Mingaleva, Zhanna ; Tomić, Duško (ur.).
Beograd, 2017. str. 469-477 (predavanje, međunarodna recenzija, cjeloviti rad (in extenso), znanstveni)
CROSBI ID: 926527 Za ispravke kontaktirajte CROSBI podršku putem web obrasca
Naslov
Cost-sensitive learning from imbalanced retail credit dataset
Autori
Oreški, Stjepan ; Oreški, Goran
Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni
Izvornik
International Scientific Conference on IT, Tourism, Economics, Management and Agriculture – ITEMA 2017
/ Nedanovski, Pece ; Filipović, Dejan ; Mingaleva, Zhanna ; Tomić, Duško - Beograd, 2017, 469-477
ISBN
978-86-80194-08-0
Skup
International Scientific Conference on IT, Tourism, Economics, Management and Agriculture – ITEMA 2017
Mjesto i datum
Budimpešta, Mađarska, 26.10.2017
Vrsta sudjelovanja
Predavanje
Vrsta recenzije
Međunarodna recenzija
Ključne riječi
genetski algoritam ; klasifikacija ; neuronska mreža ; ocjena kreditnog rizika ; neuravnoteženi skup podataka ; trošak pogrešne klasifikacije.
(genetic algorithm ; classification ; neural network ; credit risk assessment ; imbalanced datasets ; misclassification cost)
Sažetak
Cost-sensitive imbalanced data exist in many challenging real-world classification problems, where the misclassification of minority class instances is usually several times more expensive than those of the majority class. Using standard classification techniques and evaluation measures produces biased results in favor of the majority class. One of the domains that is sensitive to this type of bias is banking, especially credit risk assessment. In the present study, a new classification technique based on genetic algorithm and neural network, optimized for the cost-sensitive measure and applied to retail credit risk assessment, is created. The relative cost of misclassification, which properly accounts for different misclassification costs, is used as the primary evaluation measure. The test of the new algorithm is performed on German retail credit dataset. An empirical comparison demonstrates the potential of the new technique in terms of misclassification costs.
Izvorni jezik
Engleski
Znanstvena područja
Računarstvo, Informacijske i komunikacijske znanosti
