Loan Frauds in the Indian Banking Industry: A New Approach to Fraud Prevention Using Natural Language Processing (NLP)

Abstract

Context/Motivation

Non-identification of Early Warning Signals (EWS) or Red Flag Indicators (RFI) on time is an important reason behind the rising trend in credit frauds in the Indian banking industry. Literature suggests that for effective identification of EWS, it is not enough to have a set of EWS but it is essential to rank them and highlight the most important ones to look out for. In the Indian context, there is no ranking of EWS for credit frauds, which is a serious challenge to practicing bankers.

Design/Methodology

This paper therefore ranks the EWS for credit frauds using a novel Natural Language processing (NLP) approach and further analyses the most important EWS impacting frauds.

Findings

This paper finds that the presence of early warning signals from Diversion of Funds, Inter-Group/Concentration of Transactions, Issues in Primary/Collateral Security (COLSEC), makes it very likely that frauds would be in the high-value category.

Originality

First, this is the first Indian study which develops a ranking or scoring of either EWS/RFI on the basis of NLP tools. Secondly, we use a unique methodology for identification of EWS based on NLP techniques, which makes possible the harnessing of a rich source of data, not so far attempted.