Dr. Sie Long Kek
Universiti Tun Hussein Onn Malaysia, Malaysia
Title: Linear Quadratic Gaussian System Identification: A Data-Driven Framework from Observation to Validation
Abstract:
In the presence of stochastic disturbances and measurement noise,accurately estimating and controlling dynamic systems remains afundamental challenge in statistical modelling. This talk presents a data-driven framework for the Linear-Quadratic-Gaussian (LQG) model thatuses historical data for system identification and parameter estimation.These parameters are subsequently used to design an optimal feedbackcontrol law that regulates the model solutions and approximatesobserved historical trajectories as closely as possible. Under theseparation principle, we decouple the process into a stochastic optimalcontrol stage and a Kalman filtering stage. We detail the statisticalproperties of the filter, specifically how the state expectation and errorcovariance ensure the optimality and unbiasedness of the estimation.The framework's efficacy is demonstrated through an empirical analysisof retail sales and currency exchange rates. Validation using Mean-Squared Error (MSE) and quadratic cost analysis indicates highobservational accuracy, confirming that this data-driven LQG approachprovides a robust methodology for predicting and regulating futuresystem behaviour based on historical trends.
Keywords: Linear-Quadratic-Gaussian (LQG), System Identification, Data-DrivenControl, Kalman Filtering, Separation Principle
Biography:
Sie Long Kek, PhD, CQRM, is currently a senior lecturer in the Department of Mathematics and Statistics, Faculty of Applied Sciences and Technology, Universiti Tun Hussein Onn Malaysia (Pagoh Campus). He received his M.Sc. and Ph.D. in mathematics from Universiti Teknologi Malaysia, Johor, Malaysia, in 2002 and 2011, respectively. He was a research associate at the Curtin University of Technology in 2009 during his Ph.D. study. His research interests include optimization and control, operational research and management science, modelling and simulation, parameter estimation, Kalman filtering, and computational mathematics. He has published over 50 papers in refereed journals and six (6) book chapters. He reviews peer-reviewed research journals, including Automatica, Optimal Control, Applications and Methods, International Journal of Control, Heliyon, Journal of Industrial and Management Optimization, Measurement and Control, Hindawi Journal of Mathematics and MDPI Journal of Risk and Financial Management. He has hosted two (3) research projects supported by the Ministry of Education Malaysia. He has supervised five (5) master's and three (3) Ph.D. students. Since 2015, he is a certified quantitative risk management (CQRM) fellow. From 2021 to 2023, he was appointed head of the research focus group, Numerical Simulation and Applications (NSA).
Email: slkek@uthm.edu.my
UTHM Staf Profile: https://community.uthm.edu.my/slkek
SCOPUS ID: 35776335100
ORCID ID: 0000-0003-0425-4422
Publons ID: H-6498-2011
Google Scholar: https://scholar.google.com.my/citations?user=wWntPq4AAAAJ&hl=en