This study analyzes two geotechnical problems using Bayesian probabilistic approach, i.e., to model the effect of initial void ratio on the soil-water characteristic curve (SWCC) of unsaturated soil and to select the optimal model for the prediction of the creep behavior of soft soil under one-dimensional straining. In the study of the SWCC of unsaturated soil, a model is proposed based on the physico-empirical hypotheses and the weight-volume relationships to consider the effect of initial void ratio on the SWCCs of the same-textured soil. The relationships of the volumetric water contents and the matric suctions between two different void ratio samples are established. The relationship of their volumetric water contents depends only on their initial void ratios. In the relationship of matric suctions, an adjustment parameter (  ) is introduced as a function of void ratio, matric suction or volumetric water content to improve the predicted results. The optimal predictive models of  are determined respectively for the coarse- and fine-grained soils using Bayesian approach. The proposed model performs well on estimating the SWCC of the same-textured soil under different initial void ratios for both the coarse- and finegrained soils. The uncertainty of predictions by the proposed model is also evaluated. In the study of the creep behavior of soft soil, the transitional Markov Chain Monte Carlo (TMCMC) method and Crank-Nicolson finite difference method are adopted to simultaneously identify all model parameters in 1 elastic plastic (EP) model and 5 different elastic viscoplastic (EVP) models. The model class selection is performed respectively based on two groups of consolidation data under different loading durations using Bayesian theorem. The present probabilistic analysis is proved to be efficient in the model updating for the creep analysis involving the complex non- iii linear models and high-dimensional problem. The analyses indicate that it is crucial to use sufficient creep test data in the identification of optimal model and the associated parameters for the prediction of creep behavior of soft soil. The corresponding uncertainty of predicted deformations is evaluated by computing the 95% credibility interval (CI).