Dynamic causal mining
Causality plays a central role in human reasoning, in particular, in common human decision-making, by providing a basis for strategy selection. The main aim of the research reported in this thesis is to develop a new way to identify dynamic causal relationships between attributes of a system. The first part of the thesis introduces the development of a new data mining algorithm, called Dynamic Causal Mining (DCM), which extracts rules from data sets based on simultaneous time stamps. The rules derived can be combined into policies, which can simulate the future behaviour of systems. New rules can be added to the policies depending on the degree of accuracy. In addition, facilities to process categorical or numerical attributes directly and approaches to prune the rule set efficiently are implemented in the DCM algorithm. The second part of the thesis discusses how to improve the DCM algorithm in order to identify delay and feedback relationships. Fuzzy logic is applied to manage the rules and policies flexibly and accurately during the learning process and help the algorithm to find feasible solutions. The third part of the thesis describes the application of the suggested algorithm to a problem in the game-theoretic domain. This part concludes with the suggestion to use concept lattices as a method to represent and structure the discovered knowledge.
views in local repository
downloads in local repository
The information is available from the following content providers: