Text this: Incorporating causal notions to forecasting time series: a case study