Time: Monday 6 December, 15:00-16:30 pm

Tencent Conference Number: 117 564 951

Speaker: Associate Professor Liu Jiapeng, Xi'an Jiaotong University

Speaker's introduction.

Dr. Liu Jiapeng is an associate professor and PhD supervisor at the Research Center for Intelligent Decision Making and Machine Learning, School of Management, Xi'an Jiaotong University. His current research interests include: decision analysis, machine learning, Bayesian methods, and big data models. He has chaired the National Natural Science Foundation of China (NSFC) youth project and the National Key Research and Development Program (NKRP) sub-projects, and the Postdoctoral Science Foundation project. His research results have been published in INFORMS Journal on Computing, European Journal of Operational Research, Omega, Knowledge-based Systems, Systems Engineering Theory and Practice, Journal of Systems Engineering and other important academic journals at home and abroad. He is currently a director of the Intelligent Decision Making and Gaming Branch of the Chinese Society for Preferential Methodology and Economic Mathematics, and a member of the Data Science and Knowledge Systems Engineering Committee of the Chinese Society of Systems Engineering.

Introduction to the report.

We propose a preference learning algorithm for uncovering Decision Makers’(DMs’) contingent evaluation strategies in the context of multiple criteria sorting. We assume the preference information in the form of holistic assignment examples derived from the analysis of alternatives’ performance vectors and textual descriptions. We characterize the decision policies using a mixture of threshold-based value-driven preference models and associated latent topics. The latter serve as the stimuli underlying the contingency in decision behavior, providing a transparent and interpretable way to explore and understand DMs’ contingent preferences. Such a probabilistic model is constructed using a flexible and nonparametric Bayesian framework. The proposed method adopts a hierarchical Dirichlet process so that a group of DMs can share a countably infinite number of contingent models and topics. For all DMs, it automatically identifies the components representing their evaluation strategies adequately. The posterior is summarized using the Hamiltonian Monte Carlo sampling method. We demonstrate the method’s practical usefulness on a real-world recruitment problem considered by a Chinese IT company. We discuss the contingent models and topics and illustrate their employment for classifying the job applicants. We also compare the approach with counterparts that use just a single preference model, implement the parametric framework, or consider each DM’s preferences individually.

(Organised by: Department of Management Engineering, Centre for Research and Academic Communication)