Predicting Loan Default in P2P Lending : A comparative analysis between fsQCA and logit model
Anik, Kaysul Islam (2019)
Anik, Kaysul Islam
Åbo Akademi
2019
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2019052216700
https://urn.fi/URN:NBN:fi-fe2019052216700
Tiivistelmä
Successful identification of the potential contributors that might lead to the outcome of loan status is a significant concern in peer-to-peer (P2P) lending system. As a part of risk management, P2P platforms attempts to keep the risk levels proportional to the expected returns. Determining the risk level for the lending process requires efficient prediction methods. As a measure of prediction, different statistical tools and intelligent computing tools might serve the purpose for the lenders. But, choosing the best tool that lead to efficient prediction of the outcome is always a challenge for the them. Moreover, with the onset of technological advancement, big data also adds to the problem by making the data asymmetric, complex, and unstructured. In order to address these issues, this research aims to focus on the dichotomous property of the crisp set with the qualitative comparative analysis. As a part of this, among many intelligent computing systems, fuzzy set qualitative comparative analysis (fsQCA) is tested with statistical logistic regression. This study covered the basic theoretical backgrounds of different methods and the terminologies of P2P lending platform as a starting point. Primarily this research compared the outcome from these models and discovered that fsQCA performs slightly better than the logistic regression for prediction model development. Additionally, fsQCA adds more explainable alternatives for the financial managers which can help them to understand the interplay of the conditions towards the outcome. Conditions that can lead to the outcome of loan default are discussed where fsQCA performed successfully with higher prediction score than the default ratio of the original dataset. Compared to the default prediction model, fsQCA performed better at developing non-default prediction model in this research. For this, separate analysis is conducted based on separate outcome since the analysis also showed that fsQCA results are non-inversible to predict the occurrence of opposite outcome. Based on the outcome of fsQCA and its property of linguistic terms, it provides more opportunity to select among the best suitable alternatives that can lead to the same outcome. It is observed to be a standard method which can handle outliers, deal with complex data and utilize mixed data. This research concludes that, fsQCA can be a potential tool of risk management for the financial managers where they need to balance between different alternatives based on real-life scenarios.