Data Partitioning:Split the data into train and test. The training data should contain 80% of the full data and the test data should contain the most recent 20% of the full data.

Build a Simple, doube and triple Exponential Smoothing model [set the 'initialization_method' parameter value as 'estimated' and when using the '.fit' command, please do make sure that the parameter is optimized=True] and calculate the Root Mean Squared Error on the test data.