PREDICTION OF LIQUEFACTION OCCURRENCES BASED ON MACHINE LEARNING TECHNIQUE CONSIDERING THE DURATION OF EARTHQUAKE MOTION

Abstract

 In the 2011 Tohoku Japan Earthquake, liquefaction occurred in various areas of the eastern Japan. In the current liquefaction prediction, geotechnical survey data if employed, and it gives accurate results. However, it if difficult to apply this method to a wide area. Therefore, it if expected to establish a method which can easily evaluate the occurrence of liquefaction extensively. In this study, we aim to construct a liquefaction prediction model using support vector machine and random forests. Based on the results of the previous studies, the duration of seismic motion, which affects the occurrence of liquefaction, if newly added as an explanatory variable. In addition, the models are grouped according to the tendency of liquefaction occurrence based on the Japan Engineering Geomorphologic Classification Map.