Imodeli yengozi yekhredithi ye-AI: Ukwenza lula ukusebenza kwengozi yekhredithi

• About the Author:
• Igama lomlobi

  I-Quantumrun Foresight
• February 27, 2023

Inkinga yokumodela ingozi yezikweletu ikhungethe amabhange amashumi eminyaka. Amasistimu okufunda ngomshini nobuhlakani bokwenziwa (ML/AI) anikeza izindlela ezintsha zokuhlaziya idatha ehilelekile futhi anikeze amamodeli ashukumisayo, anembe kakhulu.

Ingqikithi yemodeli yengozi yekhredithi ye-AI

Ingcuphe yekhredithi ibhekisela engcupheni yokuthi obolekayo uzohluleka ukukhokha imali ebolekiwe, okuholela ekulahlekelweni kokugeleza kwemali kombolekisi. Ukuze uhlole futhi ulawule le ngozi, ababolekisi kufanele balinganisele izici ezifana namathuba okuzenzakalelayo (PD), ukuchayeka ngokuzenzakalelayo (EAD), kanye nokuzenzakalelayo okunikezwe ukulahlekelwa (LGD). Imihlahlandlela ye-Basel II, eyanyatheliswa ngo-2004 futhi yaqala ukusebenza ngo-2008, inikeza imithethonqubo yokulawula ubungozi bezikweletu embonini yamabhange. Ngaphansi Kwensika Yokuqala Ye-Basel II, ubungozi bekhredithi bungabalwa kusetshenziswa ukulinganisa, okusekelwe kusisekelo sangaphakathi, noma indlela esekelwe kuzilinganiso zangaphakathi ezithuthukisiwe.

Ukusetshenziswa kokuhlaziywa kwedatha kanye ne-AI/ML sekuya ngokuya kwanda ekufanekiseni ubungozi besikweletu. Izindlela zendabuko, ezifana nezindlela zezibalo namaphuzu esikweletu, zengezwe ngamasu athuthuke kakhulu angaphatha kangcono ubudlelwano obungewona omugqa futhi akhombe izici ezicashile kudatha. Ukuboleka kwabathengi, imininingwane yabantu, ezezimali, ezokuqashwa, kanye nedatha yokuziphatha konke kungafakwa kumamodeli ukuze kuthuthukiswe amandla abo okuqagela. Ekubolekeni kwebhizinisi, lapho kungekho maphuzu esikweletu ajwayelekile, ababolekisi bangasebenzisa amamethrikhi enzuzo yebhizinisi ukuhlola ukufaneleka kwekhredithi. Izindlela zokufunda zomshini zingaphinda zisetshenziselwe ukunciphisa ubukhulu ukuze kwakhiwe amamodeli anembe kakhulu.

Umthelela ophazamisayo

Ngokusetshenziswa kwemodeli yengozi yekhredithi ye-AI, ukubolekwa kwabathengi kanye nebhizinisi kungasebenzisa amamodeli okuboleka anembe kakhudlwana nanamandla. Lawa mamodeli anikeza ababolekisi ukuhlolwa okungcono kwababolekayo futhi avumele imakethe yokuboleka enempilo. Leli su lizuzisa ababolekisi bebhizinisi, njengoba amabhizinisi amancane bengenakho ibhentshimakhi yokwahlulela ukufaneleka kwawo ngesikweletu ngendlela efanayo imiphumela evamile yekhredithi esebenza ngayo kubathengi.

Olunye uhlelo olungase lube khona lwe-AI ekufaniseni ubungozi bekhredithi ukusebenzisa ukucubungula kolimi lwemvelo (NLP) ukuhlaziya idatha engahlelekile, njengemibiko yenkampani nama-athikili ezindaba, ukuze kukhishwe ulwazi olufanele futhi kuzuze ukuqonda okujulile kwesimo sezimali somboleki. Okunye okungase kusetshenziswe ukuqaliswa kwe-AI echazekayo (i-XAI), enganikeza ukuqonda mayelana nenqubo yokuthatha izinqumo yemodeli futhi ithuthukise ukucaca nokuziphendulela. Kodwa-ke, ukusebenzisa i-AI ekufanekiseni ubungozi bekhredithi nakho kuphakamisa ukukhathazeka kokuziphatha, njengokuchema okungaba khona kudatha esetshenziselwa ukuqeqesha amamodeli kanye nesidingo sokwenza izinqumo ezinomthwalo wemfanelo nezichazekayo.

Isibonelo senkampani ehlola ukusetshenziswa kwe-AI engozini yesikweletu i-Spin Analytics. Isiqalisi sisebenzisa i-AI ukubhala ngokuzenzakalelayo imibiko yokulawula ubungozi bezikweletu ezikhungweni zezimali. Inkundla yenkampani, i-RiskRobot, isiza amabhange ukuthi ahlanganise, ahlanganise, futhi ahlanze idatha ngaphambi kokuyicubungula ukuze aqinisekise ukuthobela imithetho ezifundeni ezihlukene, njenge-US naseYurophu. Iphinde ibhale imibiko enemininingwane yabalawuli ukuze kuqinisekiswe ukunemba. Ukubhala le mibiko ngokuvamile kuthatha izinyanga ezingu-6-9, kodwa i-Spin Analytics ithi inganciphisa leso sikhathi sibe ngaphansi kwamaviki amabili. 

Izicelo zemodeli yengozi yekhredithi ye-AI

Ezinye izinhlelo zokusebenza zemodeli yengozi yesikweletu ye-AI zingabandakanya:

• Amabhange asebenzisa i-AI ekufaniseni ubungozi bezikweletu ukuze anciphise kakhulu isikhathi nomzamo odingekayo ukukhiqiza imibiko enemininingwane, okuvumela izikhungo zezezimali ukuthi ziqalise imikhiqizo emisha ngokushesha nangezindleko eziphansi.
• Amasistimu anamandla e-AI asetshenziselwa ukuhlaziya amanani amakhulu edatha ngokushesha nangokunembe kakhudlwana kunabantu, okungase kuholele ekuhlolweni kwengozi okunembe kakhudlwana.
• Abantu abaningi 'abangenawo amabhange' noma 'abangenawo amabhange aphansi' kanye namabhizinisi emhlabeni osathuthuka athola ukufinyelela ezinsizeni zezezimali njengoba la mathuluzi anoveli emodeli yengozi yesikweletu angasetshenziswa ukuze kubonakale futhi kusetshenziswe amaphuzu ayisisekelo esikweletu kule makethe engasikiwe.
• Abahlaziyi abangabantu abaqeqeshelwa ukusebenzisa amathuluzi asekelwe ku-AI ukunciphisa ubungozi bamaphutha.
• Amasistimu obuhlakani bokwenziwa asetshenziswa ukuze kutholwe amaphethini omsebenzi wokukhwabanisa, ukusiza izikhungo zezezimali zehlise ubungozi bokubolekwa imali okuwumgunyathi noma izicelo zezikweletu.
• Ama-algorithms okufunda ngomshini aqeqeshwa kudatha yomlando ukuze enze izibikezelo mayelana nengcuphe yesikhathi esizayo, okuvumela izikhungo zezezimali ukuthi zilawule ngokuqhubekayo ukuchayeka okungaba khona engozini.

Imibuzo ongaphawula ngayo

• Iyiphi imethrikhi okholelwa ukuthi amabhizinisi kufanele ayisebenzise ukuze alinganise ukufaneleka kwawo kwekhredithi?
• Uyibona kanjani ngeso lengqondo i-AI ishintsha indima yabahlaziyi bengozi yezikweletu zabantu esikhathini esizayo?