Time Series Clustering based on forecast densities: Enhance TSclust package by adding automatic forecasting functionalities in the forecast package #13
The TSclust R package for time series clustering supports several methods in this family, but its functionality and ease of use could be greatly expanded by using some methods in the forecast R package.
A simple TO-DO list for this project would be:
Discuss the best way of integrating forecast, e.g. by wrapping some methods, accepting forecast objects,...
One way of defining similarity between two time series is by comparing their respective forecast densities at a given horizon. See Alonso et al., (2006) for an extended explanation of the underlying principle.
The TSclust R package for time series clustering supports several methods in this family, but its functionality and ease of use could be greatly expanded by using some methods in the forecast R package.
A simple TO-DO list for this project would be: