Abstract: Commercial bank liquidity management has been seriously concerned by bank managers.Operational research planning is the main methods of management of bank liquidity.Some parameters in model are assumed deterministic or random variables whose distributions are decided by historical data regression.But in reality,these parameters are of ten uncertain,and there are no statistical basis for their probability distribution,leading to results contrary to the scientific decision making.This paper presents an improved method to solve this problem that firstly the fuzzy set so funcertain parameters should be established,then converted into the definite distributions which be embedded in the dy namic programming model,which builds the multi stage dynamic linear prog ramming with simple recourse.Through the change of partial parameters in model the model gives different fluidity management strategy choices for different risk at titude treasures.

Key words: liquidity risk, dynamic programming, fuzzy qualitative constraints