[Objective] To study the landslide displacement prediction model, in order to provide a scientific basis for government departments to implement more reliable disaster prevention and control decisions. [Methods] An iterative unscented particle filter (IUPF) method based on singular value decomposition (SVD) constrain was proposed to establish a landslide displacement prediction model based on displacement parameters. [Results] The SVD method was effectively improved the robustness of Sigma point calculation in the unscented particle filtering method, thereby improving the prediction accuracy of the algorithm and making a more accurate prediction of the landslide stability trend. The algorithm was applied to the application and analysis of the data related to the monitoring project of the Paomashan landslide in Zhenjiang City and the landslide monitoring project on the south side of the Yuhua Interchange in Beijing-Hong Kong-Macao Expressway. [Conclusion] The prediction in landslide displacement with the unscented particle filtering algorithm with SVD constraint can be more accurate, and the predicted data can reflect the deformation trend of the landslide more accurately.