The area under the Rock curve is a common criterion for evaluating the classification performance of biomarkers. In practice, a biomarker has limited classification power, so to improve the classification performance, we are interested in combining the values ​​of biomarkers linearly and nonlinearly. In this study, while introducing the types of loss functions, the Ramp AUC method and some of its features are introduced as a statistical model based on the area under the Rock curve. This model is presented to combine biomarkers linearly or nonlinearly with the aim of improving the classification performance and minimizing the empirical loss function based on the Ramp AUC loss function. As an applied example, this study used data from 378 diabetic patients referred to the Ardabil and Tabriz diabetes centers in 2014-2015. The RAUC method was used to classify diabetic patients in terms of functional limitation status based on demographic and clinical biomarkers. Model validation was performed using the training and testing method. Based on the results of the experimental group, the area under the curve values ​​obtained for the RAUC model with linear combinations of biomarkers in the form of a linear kernel are 0.81 and with a radial basis function kernel are 1.00. The results indicate the presence of a strong nonlinear pattern in the data, such that nonlinear combinations of biomarkers have higher classification performance than linear combinations.