Track: Data Analytics and Big Data

Abstract

In the credit acceptance process, the financial institutions analyze the borrowers’ creditworthiness through their demographic data based on the 5C principle; character, capacity, conditions, capital and collateral. However, the legacy credit scoring methods have drawbacks, including not having an excellent credit reputation as it is limited to the structural nature of demographic data. This study proposes a new credit scoring approach by adding two different aspects through social media unstructured data; content and network aspects. The content aspect considers creditworthiness by assessing borrowers’ posts content including opinions and conversation in social media, while the network aspects considers borrowers’ connectivity to their social community. We construct a credit scoring model by combining the two social media elements and the demographic element. We propose a new model of credit scoring to better represent the quality of borrowers’ characteristics and behavior. The data is collected from LinkedIn which suitable to represent the professional network. The proposed model has been verified through expert judgment including the credit providers and has been simulated through a machine learning approach to automate credit acceptance decisions.