Information distribution method relevant in fuzzy reasoning

Abstract This paper introduces a new approach in formulating a fuzzy relation which differs from classical or set-valued statistics in that the statistic work is done on information gains accumulated from actual observation records at some controlling points rather than on the records themselves. Benefit comes from the case where basic observation records are scarce and it is hard to carry out statistics. A demonstration based on error analysis induced from both approaches is produced numerically on a computer. Results show that this method is effective for linear or quadratic distribution functions.