Run the Python code again in Colab. No need to run code snippets that are related to plotly. Because the plots don’t seem to appear in Colab.

Let us focus on the forecasting of the Paris demand. In the existing code, we use 61 days to evaluate the forecast performance of a model with and without holiday effects. It turns out that the model with holiday effects dominates the initial model, and this makes intuitively sense because the model becomes more flexible. To make this point stronger, do the forecast comparison again but rather for 61 days we ask you to change this to the 90 days. 

What is the mean absolute error with holidays ?