Modulai developed an interpretable, robust time-series forecasting solution for a major e-commerce checkout provider. By decomposing a complex credit-risk challenge into distinct ML subproblems, the system produced forecasts that enabled earlier detection of high-risk customers and reduced credit losses.
Stats
50%
High-risk detection rate
75%
Precision threshold
~0.5%
High-risk customer share
Challenge
Forecasting future balances is highly valuable for financial institutions managing customer payments. Insight into how a customer's balance is likely to evolve allows proactive action rather than reactive intervention, which is critical for effective risk management. Financial time series are often noisy, volatile, and influenced by irregular events, making reliable and interpretable forecasting difficult.
Solution
The solution combines four machine learning models. First, two separate models predict the total incoming and outgoing amounts over the forecasting period without assigning transactions to specific days. Then, two additional models forecast how these volumes are distributed across days within the period, creating transaction time series. These forecasts are combined with the company's balance policies to compute predicted customer balances over time. This modular approach constrains the final forecast by total transaction volumes, improving accuracy while maintaining interpretability.
Tools
The solution uses a modular ensemble of four machine learning models for volume prediction and temporal distribution. Each step can be traced back to individual models, making the results easier to explain, validate, and act upon in a risk management context. Prior domain knowledge is encoded into the modeling framework to reduce complexity while maintaining strong predictive performance.
Value created
The client was able to identify high-risk customers with high accuracy. Although only about 0.5% of customers were classified as high risk, 50% of them were identified at a 75% precision threshold, enabling more accurate precautionary actions.
Learn more
