2026 Volume 21 Issue 2 Pages 99-108
Ground investigations at construction sites are limited due to time and cost constraints, and the number of in-situ tests and mechanical test results is even more scarce, leading to soil modeling based on limited datasets. Therefore, this study aimed to secure soil data from existing databases and estimate unobserved parameters. First, to secure soil data, existing soil databases were utilized, and datasets similar to the construction site were extracted using distance calculation methods. Next, for estimating unobserved parameters, linear Bayesian regression was used to calculate estimates and 95% prediction intervals. By constructing linear Bayesian regression using data similar to the construction site, estimation results equivalent to the correct values were obtained. On the other hand, the 95% prediction interval not only indicates the reliability of the estimates but also suggests the need to review the dataset or regression model if the range of the confidence interval is large.
