In underground engineering, in-situ stress is a significant load. During the investigation phase, in-situ stress measurements are typically conducted to understand its overall distribution. However, due to limitations in field testing conditions or economic factors, stress tests are generally performed at a few representative locations. Then, the in-situ stress distribution across the entire project area is inferred through stress inversion, providing a basis for engineering design and construction. Currently, there are several methods for in-situ stress inversion, with two being the most commonly used. One method is based on constructing different finite element models of in-situ stress components and then combining them. This approach offers high accuracy but requires significant modeling effort. The other method directly uses in-situ stress test data and employs a purely mathematical statistical regression approach for stress inversion. This method has the advantages of simplicity and a smaller workload, but it cannot account for the effects of major geological structures, such as faults, on in-situ stress. This paper compares and analyzes the differences between these two methods based on several actual projects. Starting from the basic characteristics of in-situ stress, it proposes a new idea for statistical regression inversion. The results show that using the statistical regression inversion method results in a deviation of less than 10% compared to the finite element inversion results. This suggests that the inversion results can be used for engineering design and are especially suitable for situations where finite element in-situ stress inversion has not yet been conducted in the early stages of a project.