Abstract: The occurrence of open-pit mine landslides is influenced by multiple factors such as geological conditions, meteorological conditions, and human mining activities. However, existing studies focus on landslide early warning under a few dominant inducing factors, without considering complex dynamic factors such as blasting vibration, excavation unloading, and rock mass structure. In scenarios with high-dimensional monitoring data, existing early warning methods have evident limitations in terms of general applicability. To address the above problems, this study proposed an early warning method for open-pit mine landslides, integrating Principal Component Analysis (PCA), G1-order Relation Analysis Method (G1), and Extension Theory (ET). Firstly, monthly displacement, internal friction angle, cohesion, effective rainfall, water content, mining slope angle, structural plane dip difference, and mining disturbance rate were selected as early warning indicators. The early warning levels for open-pit mine landslides were classified into four levels: blue (low risk), yellow (general risk), orange (high risk), and red (extremely high risk). Secondly, PCA was used to reduce the dimensionality of the monitoring data corresponding to the indicators, extract principal component information, and determine the comprehensive importance ranking of the indicators. Then, the G1 method was applied to determine the importance ratio between adjacent indicators and calculate the weights of the early warning indicators. Finally, an element model was constructed based on ET, and the single-indicator correlation degrees were calculated through classical domain, segment domain element, and the element representing the object under evaluation, followed by weighted calculation of the comprehensive correlation degree. The early warning level was determined according to the maximum correlation degree principle. The application results showed that the early warning level of open-pit mine landslide obtained by this method was blue, which was consistent with the actual slope condition.