ANN and SVM Based modeling of stabilized soils

Flexible pavement structure consists of a different layer system. The layers are designed with different materials to reduce the stress on the subgrade. The usage of industrial products in pavement construction has gained a lot of attention in the last few decades. The utilization of the industrial products in strengthening the base layer is showing better performance in Low Volume Roads. Glass is a pozzolanic material when blended with soil it gives additional strength. With this background, this paper is aimed to provide an optimum percentage of glass powder as a soil stabilizer to improve the performance of the base layer and a suitable machine learning technique is also identified to predict the strength of the soil. The two types of machine learning techniques used in this study are SVM and ANN. Various performance measures like RMSE, R-Squared, MSE, MAPE, MAE, and Chi- square are calculated and compared between ANN and SVM. From this study, it can be concluded that ANN is the best fit model than SVM to predict the strength parameters of the soil. The optimum percentage of glass powder as soil stabilizer was found to be 15%.