Risk evaluation is a key management step employed to avoid or prevent investment risk by appraising the risk level of a certain investment project before its execution. An artificial intelligence method is an effective choice of investment risk management. As a component of the artificial intelligence method, the artificial neural network(ANN)is dealt with in this paper. The feasibility of the ANN as an appraising tool against investment risk is discussed. Thus, the ANN model is set up, and its structure is carefully designed．In the ANN mod eI as risk evaluation．the inputs of the network should be the indices that desaribe or may cause the investment risks，and then the output wilI show the risk Ieve1．On the basis of the investment risk distinction．the input units and the output units are defined．The concealed Iayers and units are also determined by making use of the ANN theories and method s．In addition，the relationship between the training frequency of the network and the precision of the output has been studied and the best training frequency thus obtained should be 200 thousand times．So ma ny times of studying 41 samples with the network have resulted in the distribution of the weights among the points of the mode1．This modeI has been applied to the rea1 cases of some investment projects to evaluate their respective risk 1evels．and the results have proved satisfactory．