Talk to be given at the International Symposium on Forecasting, Rotterdam. 1 July 2014
This is a short course I am giving at Humboldt University, Berlin, 24-25 June 2014.
Abstract: Functional time series are curves that are observed sequentially in time, one curve being observed in each time period. In demography, examples include curves formed by annual death rates as a function of age, or annual fertility rates as a function of age. In finance, functional time series can occur in the form of bond yield curves, for example, with each curve being the yield of a bond as a function of the maturity of a bond. I will discuss methods for describing, modelling and forecasting such functional time series data.
I am giving a two-part seminar at the Energy Forum, Valais/Wallis, Switzerland, on 17 June 2014.
Abstract: Electricity demand forecasting plays an important role in short-term load allocation and long-term planning for future generation facilities and transmission augmentation. It is a challenging problem because of the different uncertainties including underlying population growth, changing technology, economic conditions, prevailing weather conditions (and the timing of those conditions), as well as the general randomness inherent in individual usage. It is also subject to some known calendar effects due to the time of day, day of week, time of year, and public holidays. But the most challenging part is that we often want to forecast the peak demand rather than the average demand. Consequently, it is necessary to adopt a probabilistic view of potential peak demand levels in order to evaluate and hedge the financial risk accrued by demand variability and forecasting uncertainty.
Alexander Dokumentov and Rob J Hyndman
Abstract: We propose a new generic method ROPES (Regularized Optimization for Prediction and Estimation with Sparse data) for decomposing, smoothing and forecasting two-dimensional sparse data. In some ways, ROPES is similar to Ridge Regression, the LASSO, Principal Component Analysis (PCA) and Maximum-Margin Matrix Factorisation (MMMF). Using this new approach, we propose a practical method of forecasting mortality rates, as well as a new method for interpolating and extrapolating sparse longitudinal data. We also show how to calculate prediction intervals for the resulting estimates.
Rob J Hyndman, Alan Lee & Earo Wang
Abstract: We describe some fast algorithms for reconciling large collections of time series forecasts with aggregation constraints. The constraints arise due to the need for forecasts of collections of time series with hierarchical or grouped structures to add up in the same manner as the observed time series. We show that the least squares approach to reconciling hierarchical forecasts can be extended to more general non-hierarchical groups of time series, and that the computations can be handled efficiently by exploiting the structure of the associated design matrix. Our algorithms will reconcile hierarchical forecasts with hierarchies of unlimited size, making forecast reconciliation feasible in business applications involving very large numbers of time series.