Estimation of Discharge and End Depth in Trapezoidal Channel by Support Vector Machines

Abstract

This paper presents the results of an application of support vector machines based modelling technique (radial based kernel and polynomial kernel) to determine discharge and end-depth of a free overfall occurring over a smooth trapezoidal channel with positive, horizontal or zero and negative bottom slopes. The data used in this study are taken from the earlier published work reported in the literature (Ahmad 2001). The results of the study indicate that the radial based function and polynomial kernels support vector machines modelling technique can be used effectively for predicting the discharge and the end depth for a trapezoidal shaped channel with different slopes as compared to the empirical relations suggested by Ahmad (2001); Gupta et al. (1993) and a back propagation neural network technique. The predicted values of both discharge and end depth compared well to the results obtained by using empirical relations derived in previous studies as well as with a back propagation neural network model. In case of discharge prediction, correlation coefficient was more than 0.995 with all three different slopes, while it was more than 0.996 in predicting the end depth using radial based kernel of support vector machines algorithm. Thus, suggesting the application and usefulness of this technique in predicting the discharge as well as end depth in the trapezoidal shaped channel as an alternative to the empirical relations and neural network algorithm. Further, a smaller computational time is an added advantage of using support vector machines in comparison to the neural network classifier, as observed in the present study.