Abstarct: Effective detection of change point is one of the important issues that researchers, in the field of quality control, are dealing with. Most of these studies have been done by maximum likelihood estimation (MLE) or machine learning (ML) methods. Deficiency of MLE method is that, the distribution of the underlying process must be known, whereas it is not so in practice. Also, deficiency of ML method is that, when we have more input variables, computational cost is too much. In this study, we use a two-stage hybrid scheme to detect the change point of a multivariate process. In the first stage, statistical models are used to reduce the dimension of variable space, and in the second stage, a ML method is used to provide a basis for detection of change point. The statistical models are logistic regression (LR) and multivariate adaptive regression spline (MARS). These two methods provide a reduced set of inputs for ML method of support vector machine (SVM). Thereafter, an effective approach is used on the SVM output to detect the change point. Our achievements show that the hybrid scheme has much better performance than the case where the SVM is applied on the whole set of variables.