Our client is a leading North American Railcar Services Company. The client’s Lease Team wanted to understand the optimal lease rates to charge to clients and a way to account for market and macroeconomic factors impacting lease rates.
The main objective was to build an analytical model and tool to forecast lease rates for different types of railcars using historical and external data.
- Insufficient data for certain car types
- Lack of clarity on how to determine correct coefficient directionality for modeling
- Utilized internal data (lease history, Rail Car Type Data, Car Fleet Data – Utilization and Momentum, customer information, maintenance curve), external data (macro factor, competition/industry performance), and external data sources like EIA, USDA, and FRED as input data for model development
- Developed hypothesis associated with lease history, rail car type, and macro-economic factor (crude oil price, GDP, etc.) for function approximation
- Performed feature engineering to identify 15 key variables out of 300 variables
- Developed lease forecasting ensemble model using time series and XGBoost algorithm
- Identified lease rate range for each car Type for selected quarter/s
- Provided lease rate prediction range for each of the 20 rail car types
- Developed a Tableau dashboard for the business to understand lease rate predictions at a quarterly level by easily modifying underlying key market factors