Author_Institution :
Center for Army Anal., Fort Belvoir, VA, USA
Abstract :
Bayesian classifiers are effective methods for pattern classification, although their assumptions on the belief structure among attributes are not always justified. In this paper, we introduce a new classification method based on the possibility measure, which does not require a precise belief model and, in a sense, it includes the Bayesian classifiers as special cases. This new classification method uses the fuzzy operators to aggregate attributes information (evidence) and it is referred to as fuzzy analysis of statistical evidence (FASE). FASE has several nice properties. It is noise tolerant, it can handle missing values with ease, and it can extract statistical patterns from the data and represent them by knowledge of beliefs, which, in turn, are propositions for an expert system. Thus, from pattern classification to expert systems, FASE provides a linkage from inductive reasoning to deductive reasoning
Keywords :
belief maintenance; expert systems; fuzzy set theory; inference mechanisms; learning (artificial intelligence); pattern classification; possibility theory; statistical analysis; FASE; belief maintenance; deductive reasoning; expert systems; fuzzy set theory; knowledge discovery; knowledge representation; machine learning; pattern classification; possibility measures; statistical evidence; Aggregates; Bayesian methods; Data mining; Expert systems; Fuzzy set theory; Information analysis; Machine learning; Machine learning algorithms; Pattern classification; Probability;
