Robust support vector regression with generic quadratic nonconvex ε-insensitive loss

•Generic non-convex quadratic ε-insensitive loss function is proposed.•New loss is a generalization of Huber loss, ramp least squares loss, and ε-insensitive loss.•Robustification parameter is introduced to regular the robustness in the loss.•Support vector regression method with this loss in presented.•Experimental results demonstrate the robustness and generalization ability of our method.