Development of artificial neural network based standalone graphical user interface application for prediction of hydraulic conductivity and compaction characteristics of lateritic soils.

The suitability of laterite for engineering construction depends largely on its hydraulic conductivity and compaction characteristics; thus, determining these characteristics is highly important. However, the tests to determine these characteristics are costly and cumbersome. To overcome this limitation, this research developed ANN based standalone graphical user interface (GUI) application to enhance the prediction of lateritic soils’ hydraulic conductivity, maximum dry density and optimum moisture content, from indices: specific gravity, liquid limit, plasticity index, linear shrinkage and fine content. To achieve this goal, laboratory tests were conducted on three hundred lateritic soil samples and the experimental datasets obtained were divided into a model dataset comprising two hundred and forty data points, which were used to develop predictive models using artificial neural network (ANN), and a checking dataset with sixty data points, which were used to validate the application. The developed ANN models were transformed into mathematical equations, which were embedded in graphical user interface application to predict hydraulic conductivity and maximum dry density and optimum moisture content. The performance of the developed ANN based standalone GUI application was appraised and validated against prominent regression-based models in the literature using various prediction performance metrics; based on these metrics, the ANN based standalone GUI application outperformed the existing regression-based models by presenting lower RMSE, MAPE and MAE values and higher R2 values