Abstract: Accurately determining the standard values of soil and rock parameters is a crucial prerequisite for reliability analysis and design in geotechnical engineering. In practical engineering,the standard value of shear strength parameters is typically estimated using classical statistical methods based on experimental results from multiple site samples. However,this process can introduce errors that affect the accuracy of the determined standard values. Based on extensive investigation data from the Ningbo Railway Transit project,the study focuses on Holocene marine silt clay in the fluid-plastic state,one of the soft soil layers that has the greatest impact on construction. A total of 352 sets of investigation data from the Ningbo rail transit project were collected and processed to select the silt clay parameters. After eliminating certain data,statistical analysis was performed,and a machine learning-based method for determining sample shear strength parameters was proposed. The model's predictive performance achieved R²=0.664 for cohesion and R²=0.818 for the internal friction angle. The method combines Bayesian theory with the maximum posterior estimation of probability characteristics to optimize the determination process. The standard values of soil and rock shear strength parameters were then determined by applying this method to field data from a regional project. The resulting values are based on a clear and objective analysis of the data.