Subjective information of human are widely utilized in every sphere of engineering. However, subjective data can not be handled in the same way as objective data, if they are not quantified. As the result, subjective data are not introduced in numerical analysis, usually. In this study, fuzzy membership functions are employed in order to quantify subjective classification information. A probabilistic model which can determine the shape of membership function from fuzzy data is proposed. Further, for the utilization of subjective information in reliability engineering, a recursive method to estimate probability density function from subjective classified data is developed. This probability density function can be used in the reliability analysis in the same way as the ordinary probability density function since all the calculation is carried out based on the probabilistic methods.