A novel ensemble model based on GMDH-type neural network for the prediction of CPT-based soil liquefaction

Abstract

This study presents a novel ensemble group method of data handling (EGMDH) model based on classification for the prediction of liquefaction potential of soils. Liquefaction is one of the most complex problems in geotechnical earthquake engineering. The database used in this study consists of 212 CPT-based field records from eight major earthquakes. The input parameters are selected as cone tip resistance, total and effective stress, penetration depth, max peak horizontal acceleration and earthquake magnitude for the prediction models. The proposed EGMDH model results were also compared to the other classifier models, particularly the results of the group method of data handling (GMDH) model. The results of this study indicated that the proposed EGMDH model has achieved more successful results on the prediction of the liquefaction potential of soils compared to the other classifier models by improving the prediction performance of the GMDH model.