Abstract: When the traditional association rules are applied to the monitoring and early warning of geotechnical engineering, the machine learning has poor real-time performance and high redundancy of association rules. Aiming at the real-time and logic requirements of the association rules in the case of massive monitoring data of landslides, a fore-part and rear-part parallel FP-growth (FRPFP) algorithm is proposed. Through the statistical classification of landslide disaster factors from Fengjie to Jiangjin of Three Gorges reservoir, 7 basic factors and 4 induced factors are set as the front set, and the displacement parameters at the monitoring points at front, middle and rear edges of the landslide are set as the rear set. In addition, the monitoring data of 25 landslides in the study area for 11 years are collected. Based on the FRPFP algorithm, a large data system of intelligent landslide monitoring and warning based on the association rules is established. The three functions, hazard criterion mining of regional landslides, hazard criterion mining of typical landslides and occurrence analysis, and mining of landslides, are designed, and implemented under the large data-distributed processing platform Spark. The engineering verification shows that the proposed model has good real-time performance and logical rules. It is used to predict and analyze the stability of the landslide on the bank, which provides a new way of thinking for identifying the failure mechanism of the bank landslides and improving the forecast level.