Statistics Seminars: Using Mixture Models for Prediction from Time Series, with Application to Energy Use Data
13 March 2017 14:30 in CM221
Locally weighted mixture models are used for predictions which have been employed for analysing the trends of time series data. Estimation of these models are achieved through a kernel-weighted version of the EM-algorithm,using exponential kernels with different bandwidths which have a much bigger effect on the accuracy of the forecast than the choice of kernel. By modelling a mixture of local regressions at a target time point but with different bandwidths, the estimated mixture probabilities are informative for the amount of information available in the data set at the scale of resolution corresponding to each bandwidth. Nadaraya-Watson and local linear estimators are used to carry out the localized estimation step. Then, several approaches for many step ahead predictions at a time point from these models are investigated with different bandwidths. Real data are provided including data on energy use of some countries from 1971 to 2011.
Contact firstname.lastname@example.org for more information