The primary challenge for our client was to mitigate the risk of loan defaults while extending their loan products to the unbanked sectors in APAC. As unbanked customers do not have a formal financial track record and credit history, normal credit scoring mechanisms cannot be used to appraise loans for them. Credit risk assessment became a difficult task in the process of loan appraisals as the customers had no previous data or credit history, based on which the assessment is usually done. The client was looking for an innovative solution to overcome these challenges in analyzing credit risk for their unbanked customers.
Our team of specialists at KANINI worked around the challenges and came up with an AI-powered predictive modeling solution to address the challenges of the client. The model used supervised learning techniques to classify defaulting customers and unsupervised learning techniques for anomaly detection in customer segments. Dataflow was used to retrieve data like customer profiles and transaction histories from various data sources and then ingest it into Big Query, which would pre-process and transform that data before finally feeding it into the AI (Artificial Intelligence) model built on Vertex AI.
This model would analyze various customer data parameters such as demographics, occupation data, transaction histories, etc. to predict the credit risk, using alternative credit scoring mechanisms.