Forecasting Stream Flow Using Support Vector Regression and M5 Model Trees
The paper presents use of two recent data driven techniques namely model tree and support vector regression to forecast stream flow at two different locations one in Narmada river basin and the other location in Krishna river basin of India. The stream flow models are developed using the previous values of measured stream flow and rainfall to forecast stream flow one day in advance. All the models (total 63) show reasonable accuracy as evident by high values of correlation coefficient, coefficient of efficiency and low value of root mean square error. Additionally scatter plots and hydrographs were also drawn to asses the model performance.