Abstract: Slope displacement data and other information at the project site were used to make the training samples and test samples, based on orthogonal experimental design and FLAC^3D numerical simulation. The potential mapping relationship between slope displacement and parameters to be back analyzed was established by the BP neural network.The BP neural network was optimized by particle swarm optimization, which then was used to search out the most likely equivalent parameters between forecast and measured displacement in the global space of BP neural network.At last, the safety factor of slope that was used to evaluate its stability was obtained by FLAC^3D.The result shows that it is feasible to carry out the back analysis of displacement, combined with BP neural network and particle swarm optimization; the stability analysis would be more exact when the parameters are gotten by the back analysis.