JPS6478327A - Inference system - Google Patents

Abstract

PURPOSE:To easily transform the ambiguity into the numeric value and to obtain the knowledge in a form that can be easily understood by a human being by transforming the fuzzy gathering data into the Dempster-Shafter probability mass distribution data and performing the Dempster-Shafter inference based on said distribution data. CONSTITUTION:The probability distribution data stored in a statistic information storing part 10 is transformed into the possibility distribution data and stored in a data storing part 14. The ambiguity information stored in an ambiguity information storing part 16 in terms of language is transformed into the possibility distribution data, i.e., the fuzzy gathering data by a transformer 18 and stored in the part 14. Then the fuzzy gathering data is approximated to a finite staircase function by a transformer 20 and then transformed into the Dempster-Shafter probability mass distribution data, i.e., the basic probability allocation data. The probability mass distribution data obtained by the transformer 20 is stored in a data storing part 22 as the knowledge producing the probability mass distribution. This data is read out to an inference engine 24 to undergo the inference via the Dempster-Shafter logic. In such a way, the ambiguity is easily transformed into the numeric value and the knowledge is obtained in a form that can be easily understood by a human being.