Abstract: The most direct and credible method to calculate the bearing capacity of the rock-socketed pile is the static load test. But because the bearing capacity of the rock-socketed pile is very large and the expense of the experiment on rock-socketed pile is high and the experiment is very hard to go to the destroy point, so it is necessary to cut down the number of the static load test but not to reduce the precision of the resuh in the scope of the engineering. According to the former lest data, and on the basis of the BP neural network, and using the toolbox of the neural network in the Matlab, the influence to the bearing capacity of the rock-socketed pile by,several controllable element is concluded. Then the reasonable design parameters of the rock-socketed pile are confirmed. The analysis shows the practical function of BP neural network method in some degree.