Patent Co - Authorship Network Analysis CS 224 W Final Report

This project aims to model the h-index of a firm as a function of several properties of the owning organization’s patent co-authorship and citation networks. Studies have shown that the number of forward citations a patent receives strongly correlate with its market value [3], which is of great interest to potential investors. Unfortunately, future number of forward citations of particularly high-value patents can be difficult to predict, due to effects such as first mover advantage [7]. Therefore, instead of attempting to directly predict number of forward citations, we propose to predict h-index, a statistic derived from forward citations instead [4]. Additionally, while most previous studies examine only characteristics of patent citation networks as forecasters, we propose to use patent co-authorship networks. These can be generated with publicly available data from the US Patent and Trade Office (USPTO) database. Our hypothesis is that the underlying organization responsible for generating the patent is a strong predictor for future productivity. Because we wish to examine if the features of the underlying organization directly contribute to a high h-index, the interpretability of our model is a top priority.