Patent Litigation Prediction Using Machine Learning Approaches

Abstract

Over the past few years, there has been a great deal of interest in patent litigation. However, in prior studies, patent litigation prediction mainly relied on bibliographic information, while other useful data was largely neglected. To fill this research gap, this study proposes an ensemble machine learning classifier to predict patent litigation. The research datasets come from the recently released United States Patent and Trademark Office (USPTO) research datasets, including patent examination and patent assignment data. Patent litigation features are trained to build machine learning models to predict patent litigation. According to our experimental results, the litigation prediction model is capable of predicting litigation with an accuracy of 79%, significantly higher than previous studies. The patent litigation prediction model can assist firms in identifying potential lawsuits, reducing lawsuit costly damage, and developing future R&D strategies.