Uncertainty analysis of an optimum predictive neural network model

Shan, C., Abbaszadeh Shahri, A., Larsson, S., Zäll, E., (2021) Uncertainty analysis of an optimum predictive neural network model in subsurface bedrock level modeling. MLRA202, Mashine Learning and Risk Assessment, Wroclaw, 25 to 28 October 2021.

The outcome of predictive geo-engineering models include uncertainties. Considering the effect of involved uncertainties in future forecasts, this study was motivated to address such a challenge in geo-engineering projects using an adequately accurate spatial subsurface bedrock model. The optimum predictive model was captured through a designed and then developed automated artificial neural network (AMM) training scheme subjected to 1967 geotechnical soil-rock soundings in Stockholm Sweden. Evaluated UQ of the predictive bedrock levels represented different ways of comparing the true and predicted value at the same points.