Open Conference Systems, CLADAG2023

Font Size: 
Tree-based regression within a hidden Markov model framework
Rouven Michels, Timo Adam, Marius Ötting

Last modified: 2023-06-20


While tree-based regression methods are popular in practice, they miss a
time series component. We thus combine regression trees with hidden Markov models (HMMs) and construct a hybrid model that can effectively capture serial correlation and the complex dependencies between the input and output variables, while also providing interpretable results. In a case study, we demonstrate that such an approach offers a powerful and flexible tool for modeling financial data. However, the presented method can be employed in many more fields, e.g. in ecology or sports.