Description
The Genesis NGN team developed a machine learning software solution for a bank that was trying to decrease bad and doubtful debts and improve the efficiency of their loan-providing and collection processes.
Our client
Our Client is a bank that has a global presence across many important cities of the world. In addition to accepting deposits, they also provide small and large loans. Industrial clients and individuals use their services.
Bad debt issue
Giving loans or money to people is easy but recovering the money is the hard part. All lending institutions face this problem. Bad debts are a common problem in every business which is why companies and financial institutions do business with people who they can trust. Despite all the screening and precautions, bad debts still occur.
The Challenge
Old methods of screening customers for loan eligibility and giving approval for the loan to a customer are slow and not 100% accurate and there is always a chance for human error and corruption, leading to bad debts. The bank wanted a system that would reduce bad debts and increase debt collection.
The solution
The following technologies helped us develop a machine-learning program to resolve the issue of bad debts and poor debt collection
- TensorFlow
- Google Cloud ML Engine
- Apache Mahout
- PyTorch
- NET
- Oryx2
- Shogun
- Amazon Machine Learning (AML)
.
As a first step, we researched the reasons for our client’s bad debt and poor debt collection issues and found that they were giving loans to people who had very little chance of repaying them on time and some of their customers were at high risk of bankruptcy or failure in business ventures. And we did market research related to what kind of people pay their loans on time. By analyzing all the data we found in our research, we created algorithms that can identify customers who are highly likely to pay back their loans and design the repayment schedule that will ensure that they pay back the loan on time without any delay. We delivered this machine-learning system to our client after months of successful testing.
Results
- The customer screening process has become smoother
- The process of loaning becomes more efficient
- Decrease in bad and doubtful debts by 2X
- Rates of default on loans decrease
- Efficiency in the debt collection process
