An approach based on Ant Colony Optimization for the induction
of fuzzy rules is presented. Several Ant Colony Optimization algorithms are run
simultaneously, with each focusing on finding descriptive rules for a specific class.
The final outcome is a fuzzy rulebase that has been evolved so that individual
rules complement each other during the classification process. This novel approach
to fuzzy rule induction is compared against several other fuzzy rule induction
algorithms, including a fuzzy genetic algorithm and a fuzzy decision tree. The
initial findings indicate comparable or better classification accuracy, and superior
comprehensibility. This is attributed to both the strategy of evolving fuzzy rules
simultaneously, and to the individual rule discovery mechanism, the Ant Colony
Optimization heuristic. The strengths and potential of the approach, and its current
limitations, are discussed in detail.
|Name||Studies in Computational Intelligence|
|Publisher||Springer Berlin Heidelberg|