Hyperparametrization test with a genetic algorithm
Jupyter Notebook
Hyperparametrization test with a genetic algorithm
On a simple historical series (margherita pizzas sold by a restaurant) I applied the classic model of Exponential Smoothing with the default parameters and the automatic ETS of stats model. I then used a genetic algorithm to find the best hyperparameters for Exponential smoothing. On the validation set I had an appreciable improvement on the MAPE. Here is the summary graph of the results.