Support vector based method for acquiring domain specific patents

Patents classification is useful in the management and some other utilities of patent. In this paper, semi-automatic patent classification system was used to solve the problem. Thus, the classification model was built to filter some domain irrelative patents. Because of different optimization target, regression model was used instead of classification model. The goal of the system is filter more domain irrelative patents while remains more domain relative patents. The experimental results demonstrate that an ideal performance could be reached through the adjustment of threshold.