Traditional prediction models are difficult to consider the uncertainty of causality of arch dam deformation, leading to the lack of causality mechanism in the established static displacement confidence interval.To solve the difficulty, a prediction method which considers the influence of cognitive uncertainty of temperature deformation effects, was proposed for the confidence interval of high arch dams’ displacement.The dynamic time warping method was used to calculate the similarity between the measured time series of multiple temperature monitoring points on the dam body, and a criterion was established to select the measured temperature deformation factors with the minimum similarity. Multiple support vector machine models were then established,in which the orthogonal experimental sampling of selected temperature factor was used as inputs. The distribution of multi-model predicted displacements was analyzed and used to establish a dynamic changed confidence interval.Research results of the Jinping I Arch Dam indicate that the proposed methods and criteria can effectively optimize the most representative dam temperature monitoring points. The minimum similarity temperature factor-based orthogonal experiment design has small error and high modeling efficiency.The confidence interval of displacement prediction is more consistent with the causal mechanism and the optimal measured temperature factor combination of the high arch dam is dynamic.