Oct 10, 2017 06:01 PM IST | Source:

Credit scoring: How AI and machine learning can help

Disruption has finally come for the lending industry in India Fintech companies are beginning to use machine learning to evaluate the credit-worthiness of Indian SMEs and millennials.

Credit scoring: How AI and machine learning can help

Anshuman Mishra

The banking industry in India has traditionally funded corporate loans to those corporations who could navigate their labyrinthine processes. This paradigm, however, is beginning to get disrupted, thanks to the advent of consumer lending startups, armed with complex algorithms and machine learning software to replace antiquated credit rating systems.

RBI has identified 12 accounts with 25 percent (Rs 1.75 lakh crore) of bank NPAs for insolvency; now, corporates are selling their assets which never happened earlier; these are amongst the top 500 exposures in the banking system. The NPAs have been attributed to cyclicality of industry sectors and that entrepreneurs shouldn’t be hounded.

However, siphoning of funds and CBI investigations like the one being done for Vijay Mallya’s Kingfisher Airlines paves the way for banks to shift focus to retail lending. Banks having a big problem with NPAs on their balance sheet, and inability to find new players to lend to as their traditional model favours high-value secured loans to enterprises as they have easily accessible financial information and credit ratings, as well as substantial collateral. This has meant that their investment portfolios were dominated by large business houses and enterprises.

Banks have tended to look upon individuals and SMEs as lower-priority customers because of their lack of credit history and collateral. Without the information or the tools with which to evaluate the risk involved in lending to them, banks would offer individuals and SMEs credit cards and personal loans; however, the high risk involved makes it an unattractive section of the market for banks, and the high interest rates, inflexibility in payment options, and the reams of red tape make it an expensive and inaccessible option for potential customers. Only wealthier individuals and others with long and consistent credit histories and ratings could typically get personal loans, even for smaller amounts.

Fintech companies have reduced the costs of credit underwriting to find the right customer through machine learning (ML). By using more data and analysing customer default probability, the credit scoring systems are able to predict behaviour, thereby helping lenders come to a more conclusive decision based on data.

ML allows innovative work on data analysis wherein a bespoke solution is being offered to consumers. This is being done by analysing rejection data of customers who’ve approached banks which earlier was not getting captured, thereby helping in the regression analysis to predict a suitable outcome. Machine learning is used to process each customer’s application as a vector of factors. It then maps the corresponding factors to enhance the chances of lending to the customer.

This has been operationalized in fintech companies through their apps which analyses new to credit customers or complex transactions. These platforms analyze thousands of non-traditional and traditional variables such as how a customer fills out a form, how much time they spend on a site, and more to accurately score borrowers, along with vast amounts of in-house data, such as customer interaction data, payments profile, and purchase transactions.

Different sets of factors in the available information are correlated and patterns are studied by the software to increase the accuracy or probability of an event. By using more data and analysing instances of default, the software is able to help lenders come to a more conclusive decision based on the data available.

Machine learning also allows for innovative work on data analysis, providing bespoke solutions by analysing rejection data of customers who have approached banks. This data, which earlier was not being captured, helps in the regression analysis that predicts suitable outcomes, reducing the level of risk in these lending transactions and reverse engineering the decision making process, reducing the costs and time involved.

There is incredible promise and potential in creating data-driven intelligent systems to determine actuarial risk. Even as the global marketplace deals with the fluctuations in this sector, the Indian fintech revolution has taken the first step in blazing a new path, beginning to capture an untapped market and disrupting the financial sector for good.

How this will all conclude remains to be seen – what is certain is that the lending industry and market in India will never be quite the same again.

(The writer is Co-Founder and CEO of
Follow us on
Available On