A genetic fuzzy linguistic rule based approach for dynamic classifier selection in distributed data enviroments

The rapidly growing amount of data being produced in the world has become a challenging problem for decision support systems. These data are located in disparate sites while time, cost and privacy concerns makes it impossible to aggregate them into one location. Information fusion systems aim to make decisions by getting the outputs of the distributed sources. Since each source is making its local decision with its own part of the entire available data, their outputs are uncertain and sometimes unreliable. Selection of competent sources for fusion is a key stage in information fusion systems. This can be done either statically or dynamically. In this paper we propose a dynamic source selection and fusion method using a fuzzy linguistic rule based system. The system is created and optimized by means of a genetic algorithm. The proposed system has the ability to deal with the curse of dimensionality and has a human understandable structure. Also by using a dynamic selection strategy in a fuzzy rule based system, it is able to make accurate decisions with multiple sources' uncertain decisions.