Social network analytics in micro-lending

Traditionally, in credit scoring, people’s banking history is analyzed to assess their creditworthiness and to determine their reliability when paying back their loans. However, as data is continuously being generated in more volume and variety than ever before, there is foundation for new credit assessment approaches, in particular by incorporating new variables to capture borrower behavior going beyond simple repayment history. Specifically, variables that describe people’s behavior have been shown to be good predictors of creditworthiness. In industry, this is being utilized by the means of smartphone applications, which facilitate micro-lending. These applications analyze the data generated when the phone is used to decide whether the person should begranted a loan. The impact of these platforms is especially important in developing countries where large portions of the population do not have any banking history, and therefore no means of receiving a loan in the traditional way. In this study, we apply social network analytics techniques to analyze a micro-lending smartphone application dataset. We build networks in various ways to capture different components of social interaction and similarity among users. These networks are then used and featurized for building scorecards and to determine which network effects are most predictive of creditworthiness. Furthermore, we discuss how the data can be utilized for fraud detection and product adoption