【企画】リスク評価とその応用

[SS02A-06] Comparison of autocorrelation function of random components in spatial distribution estimation of geotechnical properties using Gaussian process regression

*Yukihisa Tomizawa¹, Ikumasa Yoshida¹, Yu Otake² (1. Tokyo Citi University, 2. Tohoku University)

Keywords:Gaussian process regression, Autocorrelation function, Random component, Information criterion

When the trend and random components of geoengineering properties is estimated using Gaussian process regression, various models such as Markov, binary noise, or Whittle-Matérn are candidates for autocorrelation function. The model selection is performed by the information criterion AIC or BIC, which takes into account the goodness of fit to the data and the complexity of the model. Two kinds of data measured in the cone penetration test are used for the study. The spatial distributions of one-dimensional and three-dimensional space are estimated based on the selected autocorrelation functions.