Skip to Main Content
A new approach to fuzzy rule induction from historical data is presented. The implemented system - FRANTIC - is a tested on a simple classification problem against a fuzzy tree induction algorithm, a genetic algorithm, and a numerical method for inducing fuzzy rules based on fuzzy subsethood values. The results obtained by FRANTIC indicate comparable or better classification accuracy, superior comprehensibility, and potentially more flexibility when applied to larger data sets. The impact of the knowledge representation used when generating fuzzy rules is also highlighted.