Support Vector Machines (SVMs) are widely used data mining techniques owing to their excellent abilities in solving both classification and regression problems. Various efforts have been dedicated to improve SVMs based on different scenarios of the real world problems. It has been observed that some monotonic relationships exist between attributes and the output values, and SVMs that consider such prior knowledge in some application fields obtain certain advantages. Previous studies showed that SVMs incorporating monotonicity constraints, so-called MC-SVM, could improve the performance of SVMs in some applications. In this research, we aim to explore the impacts on predictive performance of MC-SVM model with respect to constructing methods as well as different number of monotonicity constraints.