Fintech Intelligence
Resources
>
Case Study
>
Fintech Intelligence
Business Problem
Leading global investment bank, looking at large-scale accurate forecasting for product segments such as deposits inflow and outflow, securities, and loans, to maximize ROI.
Requirements
- Dynamic pricing model to improve margins by region and provide additional services.
- Identify potential outlet churn through forecasting.
Challenges
- The models need to be constantly validated against actuals.
- Fed decisions and rules have a huge impact.
Solution
- Extracted and transformed all business-critical data across regions and product verticals into an aggregated data store.
- Built multiple forecasting models across different segments.
- Strategically incorporated seasonality and external factors like rate hikes.
- Implemented ML models and Deep learning models based on product type to forecast the cash flow on monthly and yearly basis.
Business Outcomes
- 35% : Projection Accuracy.
- 60% : Resource reduction
Tech Stack
- Stats Models
- Auto-regressive Integrated
- Seasonal Auto-Regressive
- Auto AR
- ARIMA & SARIMA
- Trigonometric Seasonality
- MI Models:
- CAT & Gradient Boost