Time Series and Rough Sets

This thesis attempts to deal with the problem of time within the framework of rough sets. The rough set theory has emphasized the reduction of information necessary to acquire desired knowledge. This is particularly important when we are dealing with time. The farther back we are tracing our dependencies, the more attributes will become independent of our current decisions. We formalize approaches to reasoning with time series where the sequence of events is important, and introduce formalisms to deduce decision rules with real-time constraints. These formalisms for time sequences without real-time constraints are implemented and experimented with. A new system called 'Rough Enough', has been developed to carry out the experiments described in this thesis. Some new variants and improvements of genetic algorithms to find reducts more efficiently are introduced.