DATA DIVISON METHOD OF SAND LIQUEFACTION SAMPLES BASED ON SELF-ORGANIZING MAPS
To some extent, artificial neural networks (ANNs) had been successfully applied to the classification of sand liquefaction. In majority of which, data division was carried out on an arbitrary basis. However, the way on how data were divided may have a significant effect on model performance. In this paper, data division and its impact on ANN model were investigated through evaluating the possibility of sand liquefaction on condition of shallow foundations. And a new data division method was investigated: data division using self-organizing maps (SOMs). Results indicated that a statistical property of data in training, testing, and validation sets needs to be taken into account for ensuring optimal model performance. Obviously, the SOM clustering method was suitable for data division.
sand liquefaction seismic settlement evaluation neural network self-organizing map back propagation
Si-Si LIU Ming-Hua ZHAO
College of Civil Engineering, Hunan University, Changsha, 410082, China
国际会议
International Symposium on Geoenvironmental Engineering(国际环境岩土工程研讨会
杭州
英文
443-448
2009-09-08（万方平台首次上网日期，不代表论文的发表时间）