Mike Ka Pui SO
Professor
Department of Information Systems, Business Statistics and Operations Management
The Hong Kong University of Science and Technology
About
Mike Ka Pui So is a Professor in the Department of Information Systems, Business Statistics and Operations Management at The Hong Kong University of Science and Technology (HKUST). His research expertise includes nonlinear time series analysis, dynamic modeling of economic and financial data, Bayesian analysis, risk management, and healthcare analytics. His work has been widely disseminated through leading international academic journals.
Prof. So serves on the Editorial Board of Scientific Reports and is an Associate Editor for Econometrics and Statistics. He is also actively engaged in university–industry collaboration, having advised on numerous projects with mutual funds, stock exchanges, technology firms, healthcare organizations, and international enterprises in the areas of finance and business analytics. Currently, he is the Regional Director of the Hong Kong Chapter of the Professional Risk Managers’ International Association (PRMIA) and the Chairperson of the East Asian Outreach Committee of the International Statistical Institute (ISI).
Featured Publications
Communications Medicine, v. 5, May 2025, article number 179
Amanda M. Y. Chu, Jenny T. Y. Tsang, Sophia S. C. Chan, Lupe S. H. Chan & Mike K. P. So
Long COVID poses major public-health challenges, making effective surveillance essential. This study evaluates whether Google Trends can improve monitoring of long COVID symptoms. Using 33 search terms and 20 related topics identified from CDC and scite sources, researchers created a merged search volume (MSV) metric to address data limitations. They found four symptom topics that rise in search popularity before “long COVID,” and nine that rise afterward. MSV successfully forecasted long COVID prevalence, indicating that Google Trends–based analytics can support risk management and early detection.
Scientific Reports, v.11 (1), March, 2021, article number: 5112
MKP So, AMY Chu, A Tiwari, JNL Chan
This study introduces a global network analysis to assess COVID-19 pandemic risk by linking 164 countries based on the co-movement of newly confirmed cases. When many countries experience rising cases simultaneously, the network’s density, clustering, and assortativity indicate heightened global spread, providing early warning signals as early as late February 2020. The authors develop two indicators: a preparedness risk score for predicting spread and a severity risk score for monitoring control. Asia contributed 25–50% of preparedness risk after February 2020, while the Americas contributed about 40% in July. By December 2020, America and Europe contributed about 90% of severity risk. An online dashboard tracks these metrics.
npj Mental Health Research, v.2 (1), September 2023 , p.15
Amanda M. Y. Chu, Benson S. Y. Lam, Jenny T. Y. Tsang, Agnes Tiwari, Helina Yuk, Jacky N. L. Chan, Mike K.P. So
Family caregivers often face heavy stress that can lead to psychosocial health problems, but early detection can help prevent long-term harm. This study developed an Automatic Speech Analytics Program (ASAP) to identify caregiver stress levels using speech data. Testing with 100 Cantonese-speaking caregivers showed that ASAP could distinguish high versus low stress burden with 72% accuracy. The results demonstrate the potential of digital tools to support psychosocial assessments. Unlike traditional evaluations that require specialists and multiple interviews, ASAP offers a rapid, cost-effective first-line screening method, enabling timely referral to social workers or healthcare professionals for further assessment and treatment.
Journal of Econometrics, v.227 (1), March, 2022, p.151-167
Mike K.P. So, Thomas W.C. Chan, Amanda M.Y. Chu
This paper presents a modified method for estimating high-dimensional dynamic variance–covariance matrices using risk factor mapping to efficiently model dependence among asset returns in large portfolios. The key idea is to capture time-varying dependence through the co-movement of returns with underlying risk factors, allowing the mapping matrices themselves to change over time. Risk factors are flexibly modeled using a copula MGARCH framework. Because the number of parameters depends on the number of risk factors rather than the number of assets, the approach is highly scalable. Bayesian methods are used for parameter and risk-measure estimation. An empirical study on the Hong Kong stock market demonstrates the method’s practical effectiveness.
Contact
Inquiries regarding research, publications, or potential collaborations are welcome.