Case Study: Retail Bank achieves 92% credit score accuracy with TurinTech evoML

Improve accuracy of credit scores with AI

A retail bank approached TurinTech to address inefficiencies in its credit risk assessment process. Their existing custom-built machine learning models, which used linear regression, were unable to incorporate alternative data sources, failed to capture complex data patterns, and suffered from overfitting due to imbalanced data, resulting in suboptimal credit scoring accuracy.

Using the evoML platform, TurinTech automated the data science pipeline to develop and optimize a new model. The solution selected was a decision tree, which boosted prediction accuracy from 77% to 92% and improved prediction speed by 4x. By effectively mitigating overfitting through advanced feature engineering, evoML delivered this production-ready model in just two weeks, a fraction of the time the bank's internal efforts had previously taken.