Loading...

A Data-Driven Framework Based on Credit Risk Management for Improving the Performance of Peer-To-Peer Lending Platforms

Ghoreishi, Mohammad Reza | 2022

122 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55562 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Eshghi, Kourosh
  7. Abstract:
  8. Peoples need financial resources to improve their quality of life, and companies need them to progress and increase their comparative advantage. Borrowing is one of the most popular methods of financing. Peer-to-peer lending, also known as "marketplace" or "social" lending, is a novel form of intermediation that expunges the role of financial institutions (like banks) as an intermediary and allows entities (individuals or businesses) to raise loans directly from other entities. This lending platform can provide better quality services to borrowers (by quickly providing a simple loan application process through a transparent and flexible portal) and lenders (by managing their funding and live tracking the status of their investments).This thesis proposes a new data-driven model for platform owners to increase the number and satisfaction of their lenders directly, which also cultivates the contentment of their borrowers. The proposed model includes a four-class classification model and a new non-linear mixed integer programming model, which also discovers different management insights using association rules and counterfactual explanations. We evaluated the value and validity of the proposed model by the "LandingClub" platform dataset as a case study from three aspects: The first evaluation showed that the proposed machine learning model has a higher screening ability than the current models. The second evaluation showed that the proposed mathematical programming model is consistent with the assumptions of the literature and is applicable in reality. Finally, the third evaluation by simulating the matching process showed that the proposed data-driven model has a better sequel than the platform's current functioning.
  9. Keywords:
  10. Multiclass Classification ; Risk Assessment ; Credit Risk ; Nonlinear Mixed Integer Programming ; Many-to-Many Matching ; Peer-to-Peer Lending

 Digital Object List

 Bookmark

No TOC