Widely used in a variety of fields in society, Machine Learning discovers the rules and patterns behind data and establishes models that can reproduce them. In addition, the established models can reasonably provide unknown information. Soil properties at any locations must be estimated using the nearby soil investigations, because they are unknown unless a soil investigation is carried out. In this paper, consolidation properties of clays at any locations without soil investigations is estimated by applying Machine Learning to a soil investigation database. The Holocene clay layer at Kobe Airport, which was constructed as a man-made island in Osaka Bay, is targeted. Many soil investigations had been carried out on the target site for construction of Kobe Airport. The consolidation characteristics of Holocene clay are estimated using deep learning as machine learning, applying many soil investigations in Kobe Airport. Moreover, through the discussion of estimation accuracy on the architecture of deep learning, even simple architecture of deep learning provides estimations with small errors. The consolidation characteristics estimated by deep learning are used to simulate the consolidation settlement behaviour of the target site by the finite element method. The good agreement between the simulated consolidation settlement behaviour and the measured consolidation settlement behaviour suggests that machine learning can properly estimate the consolidation characteristics. Through these considerations, it became clear that the consolidation characteristics of soil at any point can be estimated by machine learning.