Time series forecasting
Forecasting implies identifying models that fit the historical data and using them to predict future values. When forecasting time series data, decomposition plays a very important part, helping to make predictions more accurate. The underlying assumption is that we can be more precise if we forecast each component individually, using the best-suited method, and then sum or multiply the parts (depending on whether the model is additive or multiplicative) to compute the final value.
Naïve
This is the simplest method, stating that the forecasted value is equal to the last value in the series. As mentioned previously, this is used with random walk models, where future movements are unpredictable. For example, to predict the value for the first unknown month, January 2018, using the naïve model, we can take the seasonally adjusted value from December 2017 and add (multiply) the seasonal component of the month of January:
julia> update(unemployment_data, Date(2018, 1,...
