A novel parallel cross-validated support vector machine on patent classification system

Classification, searching, monitoring, and analyzing of patent are time consumption tasks and they will significantly influence the decision-making of intellectual property on optimizing resource allocation in knowledge or technology-intensive enterprises. In the age of fundamental change, the rapidly growth of patent documents has become one of the difficult problems in developing an efficient patent classification system. A well-trained patent classification system usually needs the cross-validation procedures to improve its performance before it can be applied to the real-world applications. However, researchers lack an efficient way to perform the cross-validation procedure of huge dataset in a short time. Therefore, this study proposes a novel approach that integrates loop level parallelism (LLP) and hybrid genetic-based support vector machine (HGA-SVM) to increase the generalization ability (external validity) of support vector machine patent classification system without the problem of accompanying huge time consuming. Compared with previous approaches, in our approach, the patent calculation/classification tasks are randomly distributed into cluster computers simultaneously by parallel computing technique for greatly reducing their calculation time. The experimental results demonstrate that our approach can improve the stability of patent classification system and spend less cross-validation time than traditional methods. We believe that our proposed approach is able to help decision-makers in identifying/classifying/monitoring critical patents for assisting their strategic intellectual property decision-making process.