Abstract: Effective forecasting model for foundation pit’s deformation trend could guide construction and avoid accidents. In this article, GM (1,1), support vector machine and BP neural network were employed for single forecasting of pit deformation, and combinatorial forecasting models with fixed-weight and non-fixed weight were also established. Furthermore, R/S analysis was carried out to determine the deformation trend and verify the effectiveness of the combinatorial forecasting results. Results suggest that combinatorial forecasting could effectively improve the stability and precision of the prediction results, among which combinatorial BP neural network has the optimal results with the measured values and predicted values in good agreement. Moreover, the deformation of foundation pit will further increase and the stability has trend of weakening, in consistency with forecasting results. The research verified the effectiveness of combinatorial forecasting and R/S analysis in judging the deformation trend of foundation pit, and provided a new idea for the prediction of foundation pit deformation.

Key words: foundation pit, fixed weight combinatorial forecasting, non-fixed weight combinatorial forecasting, R/S analysis, stability analysis, trend judgment