27Nov
Seminar Calendar, Information and Innovation Management
Statistical Random Forests Modelling for Nonlinear Time Series
27 Nov 2026 | 10:30 a.m.— 11:30 a.m.
KK 1121, K. K. Leung Building, HKU
SPEAKER
Professor Zudi LU
Professor of Biostatistics
Department of Biostatistics
City University of Hong Kong
ABSTRACT
Random forest is a popular machine learning method in many applications. It heavily depends on integrating the trees constructed by the technique of bootstrap resampling of the original data. However, for the time series, it is often challenging to do this owing to the unknown underlying dependence of the data. This paper explores the development of an efficient Random Forests by Random Weights (RF-RW) methodology tailored for time series modeling. Addressing limitations of traditional random forests, the proposed approach handles dependent time series data, moving beyond the assumption of independent and identically distributed (i.i.d.) observations. The project combines theoretical advancements with practical evaluations, including simulation studies and applications to UK COVID-19 daily case modeling. Results demonstrate the superior performance of RF-RW compared to existing methods like original random forests and random forests with traditional time series resampling as well as other machine learning methods including SVMs, and LSTMs.










