The aim of this short project was to investigate the behavior of statistical time-series models and their mathematical assumptions. Main subject of this study are ARIMA and SARIMA models that were used to understand the relationship between consumption and GDP in Australia. Additionally, forecast combinations were derived to identify further modelling variants.
Furthermore, the performance of the models was compared to ETS, TBATS, Average. DR+ARIMA etc. The results indicated, that DR+ETS approaches performed best considering the given data.
- Analysis of Trend, Seasons, and Spurious Regression Issues (KPSS, ADF)
- Stationarity Analysis
- Autocorrelation and Residual Analysis
- Box-Cox Transformation
