Radar automatic target recognition is a technology that applies target recognition to traditional radar. With the rapid development of information technology and the complex battlefield situation, radar automatic target recognition technology has received intensive attention in recent years. The target recognition methods can be divided into generative models and discriminant models. The generative models focus on the description of the data distribution. The discriminant models aim to distinguish the differences between different classes data. The hybrid model which combines generative and discriminative models can enhance the recognition accuracy and contain all benefits of both these models. This paper proposed a hybrid model max-margin regularized factor analysis (MMRFA) model. MMRFA model combines the factor analysis (FA) model and latent variable support vector machine (LVSVM) in original space. The experimental results show that the MMRFA is efficiency and performs validly in low SNR situation.