Forecasting non‐seasonal time series with missing observations