SVR is a novel neural network algorithm technique based on statistical
learning theory that has received increasing attention as a
method for solving nonlinear regression estimation problems. SVR is
derived from the structural risk minimization principle to estimate a
function by minimizing an upper bound of the generalization error
[55]. It has been successfully applied in different time series prediction
problems such as production value forecasting, traffic flow prediction,
and financial time series forecasting [6,36,42,50,51].
The SVR model can be expressed as the following equation [55].