The course provides an overview of time-series based forecasts together with a theoretical basis of time series models. (Since the focus is on time series methods of forecasting, there is some overlap with the class Time Series Analysis.) To this end, we first introduce generic (and useful) concepts for time series modelling (stationarity, ergodicity, martingale difference), followed by a discussion of univariate linear processes (motivation, properties). We then extend the scope of the course to multivariate models, specifically the vector autoregressive process. Here, we also discuss estimation of autoregressive models together with basic model-based forecasting concepts. The final part of the course covers the basics of forecasting theory (loss functions, optimality, rationality) as well as practice (model-based forecasting, forecast evaluation).
- Lehrende:r: Matei Demetrescu
- Lehrende:r: Matei Demetrescu
- Lehrende:r: Fabian Schmidt