Statistical Consideration of Holocene Clay Properties Estimated by Artificial Neural Network in Kobe Airport

The spatial distribution of the consolidation properties for a seabed must be appropriately estimated to accurately predict the settlement behavior due to large-scale reclamation. In this study, an artificial neural network was applied to spatially interpolate consolidation properties such as the void ratio, compression index et al. The estimation accuracy of consolidation properties with an artificial neural network, especially natural water content, void ratio and one-dimensional compression curve, was excellent from judging based on four stochastic indexes: R², G, MARE, and SR as well as relationships between probability density and relative errors. Also, the spatial distribution of consolidation properties in the Holocene clay layer was visualized. Consequently, the availability of spatial interpolation of consolidation properties by using an artificial neural network was confirmed.