Research in this area seeks to apply machine learning methods with possible integration of other methodologies such as decision analysis and operations research to enhance decision making through development of intelligent and adaptive computer-based systems. Some recent work include:
1. Diagnostic and decision system for sepsis in critically ill children
2. Predictive modeling of hospital readmission for chronic diseases.
3. Products categorization for optimal production planning.
4. Reduction of inventories for assembly-based manufacturing
Research in this area seeks to develop all aspects of the mathematics and methodologies of decision making and risk management in sustainable and resilient complex systems. The program examines decision making and risk from various perspectives, including operational and strategic; development of techniques for sustainable decision making in conditions of uncertainty, conflict, and time critical; making sense of multiple conflicting sources of information, etc. Some recent work includes:
1. Flexibility in complex system design, planning and expansion.
2. Efficiency modeling and measurement of networked systems of system.
3. Integrated MCDM approach to sustainability and resilience in supply chain design and management.
4. Integrated MCDM Decision Support in Environmental and Water Resource Management.
Research in this area focuses on the modeling and optimization of complex large-scale systems, including the use of heuristics and meta-heuristics, and the development of computationally tractable approaches for finding optimal or near-optimal solutions. Some recent work includes:
1. Multiple junctions online traffic signal control optimization.
2. Vehicle Routing Problems with uncertain demands.
Deception and counter-deception play an important role in many warfare and conflicts. Deception involves activities designed to mislead an adversary by manipulation, distortion, or falsification of evidence to induce him to react in a manner prejudicial to his interests. This is accomplished by creating a perception in the adversary that causes him to believe in a false reality, thus enabling us to achieve our desired goals. In this project, we have developed decision support models and tools for deception and counter-deception. The methods are based on using Bayesian networks as the knowledge representation tool for situational threat assessment in scenarios involving a potential adversary. The use of Bayesian networks in situation assessment in defense and security has been well-developed and widely applied. Our methods perform deception detection by analysis of inconsistency of information in the Bayesian network. We also proposed a measure to quantify the degree to which the information or evidence might be a deception. We have also developed an approach based on the concept of mutual information from information theory to identify events that has a high value in serving as a candidate for planning a deception on the adversary and outlined the steps necessary to execute the deception plan via probabilistic reasoning in the Bayesian network. The models have been tested on two case scenarios from littoral warfare and cyber-attack.
Multiple UAV cooperation has been important due to the effectiveness of planning for a group of multiple UAVs performing a mission than planning separately for individual UAVs. However, this can be computationally difficult due to multiple UAVs being considered. In this project, we plan to develop efficient algorithms to help the path planning for such multiple cooperating UAVs. This would involve formulating the path planning problem as an appropriate optimization problem that takes into account the mission goals or targets as constraints and then solving for good quality or optimal solutions. Unlike existing approaches that simplify the problem through linearization at the expense of model accuracy, our approach would not be neglecting the nonlinear characteristics of the problem as the proposed algorithms could solve nonlinear optimization problems directly and may exploit any underlying problem structure to reduce computation time. Moreover, the proposed algorithms would consider possible real-time information updates arising from the cooperation of the UAVs, which is often lacking in existing approaches. The proposed algorithms could be integrated into an intelligent decision support system for handling the path planning of multiple cooperating UAVs. Testing of the performance of the proposed algorithms would be done with real or simulated data.
The purpose of this project is to develop adaptive intelligent systems for decision support for time-critical applications in dynamic and uncertain environments. We have completed a comprehensive literature survey and have developed systems architectures, decision models and computational algorithms for adaptive reasoning and decision making in two domains, namely (1) adaptive manufacturing job-shop systems and (2) adaptive multi-agent systems. These are domains with significant applications in the real world and in the industry. The proposed architectures and algorithms were evaluated and tested in a virtual environment using discrete events simulations developed especially for the two domains.
In the adaptive manufacturing decision support systems, the dynamic job-shop scheduling and rescheduling under uncertainty in a manufacturing environment were studied. Dynamic Bayesian Networks and Influence Diagrams construction approaches have been adopted as the main formalism in the system. This work advances the current state of the art in industrial jobs-shop scheduling and will enable the building of more effective decision support systems for industrial applications such as in the electronics and semi-conductor manufacturing sectors.
In the adaptive multi-agent decision systems, the rocks forging problem has been selected as the application problem. The system was tested for the effectiveness of the various agents' decision making rules, how information is communicated and shared among the cooperative agents, as well as the agent's learning schemes. This work advances the current state of the art in the architecting of decision support systems that has to deal with multiple players. These are applicable to the building of command & control systems in many domains such as military, maritime security, air-traffic control, etc.
This project is funded by the Defense Sciecne and Technology Agency's (DSTA) Defense Innovative Research Program and is in collaboration with Decision Support Solutions Center of DSTA. The fundamental goal of this project is to develop a set of advanced computational techniques and algorithms that facilitate the building of advanced planning systems for the allocation of resources and intelligent decision systems that are capable of responding to sensory inputs and providing optimal courses of action to decision makers in an uncertain and dynamic environment. The project will focus on the optimization of large-scale rosters that combines column generation with constraint programming, and the development of intelligent normative decision systems for automated reasoning and decision making in an uncertain and dynamic environment.
This Agency for Science, Technology & Research (A*STAR) funded-project is in collaboration with the Department of Computer Science, Department of Medicine, the National Neuroscience Institute of Singapore, Johns Hopkins Hospital of USA, and ReasonEdge Technologies. This work aims to develop a set of advanced decision engineering techniques to support effective prognostic analysis in medicine; the resulting techniques will be incorporated into a set of prototype applications that automate clinical practice guideline generation in significant and time-critical healthcare domains. Prognostic analysis is a critical part of evidence-based medicine that emphasizes the effective use of information to improve quality, reduce variation, and manage resources in healthcare procedures. The prognostic analysis illuminates the natural, as well as the expected, post-intervention course and outcome of disease processes. It plays an important role in care management tasks, including cost-effective diagnostic tests and treatment planning, prognostic prediction, pharmacoeconomic analysis, health technology and policy assessment, and clinical practice guideline generation.
The Systems Modeling & Analysis Laboratory Collaborator: Mindef, Singapore.
NUS Medical Informatics Program - Clinical Decision Systems Group. Collaborator: NUS School of Computing.
Application of Information Technology to Teaching, Tutoring and Assessment. Group members: Prof. AYC Nee (Principal Investigator), A/Prof. MA Mannan, Dr. SK Ong, A/Prof. SBG.Leng, A/Prof. Kassim, A/Prof. WT Chan, A/Prof. KL Poh.
Intelligent Multiple Criteria Decision Systems.
Intelligent Guidance for Multiple Criteria Decision Support Systems. In collaboration with Prof. M.A.Quaddus and Jie Lu, Graduate School of Business, Curtin University of Technology, Perth, Australia.