On-line Redundancy Elimination in Evolving Fuzzy Regression Models using a Fuzzy Inclusion Measure

This paper tackles the problem of complexity reduction in evolving fuzzy regression models of the Takagi-Sugeno type. The incremental model adaptation process used to evolve such models over time, often produces redundancies such as overlapping rule antecedents. We propose the use of a fuzzy inclusion measure in order to detect such redundancies as well as a procedure for merging rules that are suciently similar. Experimental studies with two high-dimensional real-world data sets provide evidence for the eectiveness of our approach; it turns out that a reduction in complexity is even accompanied by an increase in predictive accuracy.