In recent years, evaluating the patenting activity has gained a renewed interest in both growth economists and trade economists. An evolution model of the patenting activity is proposed by applying the potential support vector machine (PSVM) and the negative binomial model in this paper. The PSVM is used to features selection and negative binomial is used to evolution. The proposed method is feasible and effective by the results, and it provides a better forecast and estimate tool for the patenting activity. It also provides a novel way for the evolution design of the other engineering.