Setting forecasting model parameters using unconstrained direct search methods: An empirical evaluation

Highlights：

• We address the parameterization of exponential smoothing forecasting models.

• We empirically test derivative-free methods for models’ parameterization.

• We adapt the exponential smoothing model to handle box-constraints.

• Derivative-free methods almost equaled the performance of a gradient-based method.

• Derivative-free methods resulted faster than the gradient-based method.