Evolutionary data division methods for developing artificial neural network models in geotechnical engineering

Abstract

In recent years, artificial neural networks (ANNs) have been applied to many geotechnical engineering problems and have demonstrated some degree of success. In the majority of these applications, data division is carried out on an arbitrary basis. However, the way the data are divided can have a significant effect on model performance. In this report, the relationship between the statistical properties of training, testing and validation sets and model performance and the effect of the proportion of data used for training, testing and validation on model performance are investigated for the case study of predicting the settlement of shallow foundations on cohesionless soils. In addition, a novel approach for data division, which is based on a self-organising map, is introduced and evaluated for the above case study. The results obtained indicate that the statistical properties of the data in the training, testing and validation sets need to be taken into account to ensure that optimal model performance is achieved. The data division method introduced in this paper is found to negate the need to choose which proportion of the data to use for training, testing and validation and to ensure that each of the subsets are representative of the available data.