Usage of basic forecasting principles to predict Housing starts.
- Spliting of Data into Training, Holdout (PLS & Bates-Granger weights) & Evaluation
- Trend, Seasonality, Cycles (MA, AR, ARMA)
- ADL Models
- Model Selection (AIC & PLS)
- Correct Errors (Robust or HAC)
- Rolling Window
- h-step ahead forecasts (Iterated & Direct)
- Forecast Combination (Simple Average, Bates-Granger, Granger Ramanathan & WAIC)
- Forecast Evaluation (White Noise test, MA(h-1) errors, Mincer Zarnowitz regression & Diebold Mariano test)
Mainly models parametric relationships. Could use ML methods to such as Hybrid learning with non-parametric methods like regression trees to account for the non-linearities.