The improper fuzzy Riemann integral and its numerical integration are proposed in this paper. Under the setting of the improper fuzzy Riemann integral, we can ®nd the expectation of fuzzy random variable. The a-level set of the improper fuzzy Riemann integral is a closed interval whose end points are the improper Riemann integrals. Thus we provide a numerical method to approximate the improper fuzzy Riemann integral by invoking the Simpson's rule. We also ®t the end points (closed interval) of the a-level set of the improper fuzzy Riemann integral as polynomials with variable a in a least-squares sense. Finally, the membership function of the improper fuzzy Riemann integral can be transformed into nonlinear programming problem and be solved by any presented optimizer. Ó 1998 Elsevier Science Inc. All rights reserved. Kewwords: Closed fuzzy number; Improper fuzzy Riemann integral; Least-squares; Nonlinear programming; Simpson's rule