Title: Clustering and forecasting multiple functional time series
Authors: Chen Tang - The Australian National University (Australia) [presenting]
Han Lin Shang - Australian National University (Australia)
Yanrong Yang - The Australian National University (Australia)
Abstract: Modeling and forecasting age-specific mortality rates of multiple countries jointly could lead to improvements in long-term forecasting. Yet data that fed into joint models are often grouped according to nominal attributes, such as geographic regions, ethnic groups, and socio-economic status, which may still contain heterogeneity and thus deteriorate the forecast results. To address this, we propose a novel clustering technique to pursue homogeneity among multiple functional time series. Using functional panel data model with fixed effects, we are able to extract common features of functional time series. These common features could be decomposed into the functional time-trend and the mode of variations of functions. The proposed clustering method searches for homogeneous age-specific mortality rates of multiple countries by accounting for modes of variations and the temporal dependencies. Through simulation studies, we demonstrate that our proposed clustering technique outperforms other existing clustering methods, unless the objects are very similar. In a data analysis, we find that the clustering results of age-specific mortality rates can be explained by the combination of the aforementioned nominal attributes. We further show that our model produces better long-term forecasts than several benchmark methods for forecasting age-specific mortality rates.