Based on BAYES data processing theory, a new method of fuzzy probability is proposed for the parameter reliability of small sample. Prior distribution of fuzzy parameter is obtained by stochastic weighted method, then posterior distribution of fuzzy parameter is got by the method of conjugate distribution. In the end, the formula of fuzzy membership function is constructed, and probability of fuzzy events is solved.The study of this paper aims at the membership function which is deduced in the condition that mean is known and deviation is unknown in normal population distribution. At the same time using information of small sample of actual experiment, the study fully utilizes prior information, so, it is convincing to evaluate reliability of fuzzy parameter. It is obviously economical.