Evaluation of liquefaction potential based on CPT data using random forest

The prediction of liquefaction potential of soil due to an earthquake is an essential task in civil engineering. In this paper, random forest (RF) method is introduced and investigated for the prediction of seismic liquefaction potential of soil based on the cone penetration test data. RF has been proposed on the basis of classification and regression trees with ensemble learning strategy. The RF models were developed and validated on a relatively large dataset comprising 226 field records of liquefaction performance and cone penetration test measurements. The database contains the information about depth of potentially liquefiable soil layer (D), cone tip resistance (), sleeve friction ratio (), effective vertical stress (), total vertical stress (), maximum horizontal ground surface acceleration () and earthquake magnitude (). Two RF models (Model I and Model II) are developed for predicting the occurrence and non-occurrence of liquefaction on the basis of combination of above input parameters. The results of RF models have been compared with the available artificial neural network (ANN) and support vector machine (SVM) models. It is shown that the proposed RF models provide more accurate results than the ANN and SVM models proposed in the literature. The developed RF provides a viable tool for civil engineers to determine the liquefaction potential of soil.