Severe and multi-dimensional challenges have been posed to the construction and development of resilient cities by recurrent mountain landslides. The characterization of landslide displacement in highly permeable and recurrent mountainous regions cannot be adequately achieved by considering only rainfall magnitude and its lagged effects. The influence of multiple factors including monitoring intervals, cumulative rainfall, water levels in diversion tunnels, and average seepage volumes was taken into account to analyze landslide displacement. It was found that the deformation of the landslide exhibits significant spatial heterogeneity. The frontal zone is primarily dominated by large-magnitude horizontal displacement with a delayed response to rainfall, whereas the rear zone is characterized by relatively small vertical displacement, which is governed by a strong coupling of hydrological parameters and is susceptible to abrupt, step-like changes under specific hydrological cycles. By integrating multi-factor monitoring and analysis on-site, the average water level and seepage volume of the diversion tunnel, among other factors, were taken as dependent variables in the landslide displacement prediction study. A decision tree and gradient boosting regression tree displacement prediction model considering multiple factor monitoring was established. This fills the gap in the study of landslide displacement regarding human engineering activities and the dynamic response of groundwater. Based on this, a differentiated zoning early warning index system can be established, thereby significantly improving the accuracy and timeliness of landslide monitoring and early warning.