This study aims to utilize a self-developed, high-efficient Bayesian back analysis framework to perform Class C prediction of the embankment constructed on soft soils. In this framework, the general simplified Hypothesis B method based on a one-dimensional elastic visco-plastic (1D EVP) model and bypassing the need to solve complicated partial differential equations, is applied to perform consolidation analysis. Also, a high-efficient sampling method, Bayesian updating with structural reliability method (BUS), is employed to solve the Bayesian back analysis problem. The Class C prediction of a trial embankment constructed at Ballina, New South Wales, Australia, is conducted using the monitoring surface settlement data. The obtained results demonstrate that the accurate long-term settlement prediction at surface can be obtained at early stages, which enables cost-effective outcomes.