Grants and Contributions:

Grant or Award spanning more than one fiscal year. (2017-2018 to 2022-2023)

Two major challenges remain in many decision-making problems, too many choices and contradictive criteria. It gets worse with large, complex, unstructured, and incomplete data. The proposed research program aims to utilize machine learning methods such as rough sets, game theory and granular computing for complex and critical decision makings. In particular, I will examine ternary decision-makings when multi-criteria are involved. Applications will be drawn from intrusion detection, financial prediction, and cancer diagnoses.

The first part is to study the issues with too many choices. Rough set theory, especially probabilistic rough sets, is one of the techniques for ternary or three-way decision-makings. In real applications, one may want to reduce the number of uncertain rules to make more informed decisions. A possible solution is to lower the expectation by considering some uncertain rules as certain rules. However, this may decrease rule accuracy level and result in unacceptable consequences. Finding a proper level of trade-off is the challenge. I will examine a few probabilistic rough set models with Gini index and genetic algorithms to resolve such a problem. The key idea of these models is to use different mechanisms for evaluating certain decision rules and uncertain decision rules for an intelligent decision-making.

The second part is to examine cases when some selection criteria are contradiction. For example, generality and accuracy are two contradictive criteria. I will apply game theory to resolve the dilemma by considering measures as players to seek for balanced positions that meet the needs of these criteria. Mechanisms to find equilibriums or thresholds in games will be examined. With the preliminary study on game-theoretic rough sets, I will further examine theory aspect of the new model, and apply it to complex decision problems. Competitive and cooperative games will be examined. This will lead to a study on game-theoretic learning.

The third part is to examine multiple agents or multiple measures decision-makings. Majority, committee, and unanimous decisions are some of traditionally used strategies. Granular computing as well as game theory will be used to reach an intelligent consensus. I will apply game theory to deal with competitive or cooperative measures and granular computing to make a decision by considering different aspects of the problem. Possible applications of this method will be feature selection and Web-based decision support systems.

In summary, I will study the possibility of building an intelligent system that assistant human to make informed and wise decisions on complex problems involving multi-criteria and multiple agents. It is hoped that this will broaden our knowledge on decision support mechanisms.