<p>To investigate the influence of uncertain suspension parameters on the ride comfort of railway vehicles, a parametric vehicle dynamic model was established to calculate the Sperling comfort indices at different speeds. By employing the BP neural network, a surrogate model was constructed with 13 suspension parameters as inputs and the comfort index as the output. Then, the regional sensitivity analysis (RSA) was performed to rank the suspension parameters according to their sensitivity indicators based on the surrogate model. It was found that, for the lateral comfort index, the four most influential suspension parameters remained consistent but their orders vary at different speeds. However, for the vertical comfort index, both the important suspension parameters and their orders remained consistent despite the variation of speed. Moreover, using the regional mean ratio and regional variance ratio functions, a quantitative analysis was performed to assess how the mean and variance of the comfort index varied with range reductions of the input. The range reductions mainly include decreasing the upper boundary, increasing the lower boundary, symmetric reduction, and fixing at quantile points. The results revealed that the mean and variance of the comfort index can be effectively reduced by modifying the ranges of the important suspension parameters. This study can offer some reference and basis for the design and optimization of the suspension parameters of railway vehicles.</p>