The learning of propositional and fuzzy-logical functions andstructures has been thoroughly explored in the past years and has becomean efficient means for knowledge acquisition. Decision trees are broadlydiscussed and used, many algorithms for the learning of optimal decisiontrees are available. Publications and implementations, however, veryoften show a considerable lack of understanding of the capacities andapplicability of constructed models, and one may see many applicationswhich are developed carelessly and with little thought and come close tobeing dangerous mistakes. It is, however, possible to develop amethodology that is based on logical equations and makes maximum use ofthe existing knowledge, but avoids inadmissible generalizations andallows comprehensive knowledge engineering. The smooth transition tofuzzy-logical structures shows the efficiency of the methodology. Thepaper gives a comprehensive survey of the deficiencies of existingapproaches and demonstrates a complete solution to all of them
展开▼
