Production function is very important in the life, so choosing the reasonable mothed for estimationg parameters of production function is very necessary. Usually, Cobb-Douglas production function is multivariable nonlinear function, and most of traditional methods have some limitations in estimating parameters. A new method for estimating parameters is proposed due to the characteristics of artificial fish school algorithm such as overcoming the local optimal value and robustness. Firstly, the parameters of production function are constructed into the artificial fish and residual sum of squares is designed into the function of food consistence, secondly, the parameters of production function are estimated by the random behavior, preying behavior, swarming behavior and following behavior. The simulation studies are carried out for getting the accurate parameters of the production function. The simulation results show that this algorithm has fast optimization speed and which possesses minimum regression residual sum of squares.