Forecasting a fleet's operating expenses with machine learning
Machine learning on historical opex and its drivers predicted a tanker fleet's operating expenses 25–35% more accurately than the traditional methods.
Challenge
A tanker owner could not predict its fleet's operating expenses accurately with traditional methods, which made forecasting and budgeting unreliable.
Solution
Arundo trained machine-learning models on historical opex data and its key underlying drivers, benchmarked several algorithms, and used known future values of the drivers to forecast.
Impact
Opex predicted with a 25–35% improvement in accuracy over the traditional methods. The owner chose not to productise the solution.