Automatic Patent Classification Using Support Vector Machines And Its Applications

Abstract

Patents as an important component belonging to innovation can serve as an index in representing the technological development level in a given industry. However, agricultural biotechnology patent (ABP) data have not been updated since 2000. The major objective of this thesis is to identify ABP issued between 2001 and 2007. Apart from traditional manual identifying methods, we focus on adopting a model built on machine learning methodology. As a result, a score will be generated for each patent to indicate its possible inclusion in the agricultural biotechnology category. We select patents based on ranking their scores and following the patent classification scheme. The analysis of 9,539 ABP issued between 2001 and 2007 shows that the quantity reaches its peak value in 2001 and follows a downward trend until 2006. Based on the patent identification result, we also run several economic analyses to verify the ABP development trend and investigate their characteristics.