Modelo preditor de licenciamento de patentes acadêmicas brasileiras nos campos tecnológicos biotecnologia e farmacêutica

Abstract

Understanding which aspects guide the potential licensing determination of a patent is a central issue for entrepreneurial universities seeking to create value for society through technology transfer. This paper proposes a Random Forest (RF) approach to predict the licensing potential of Brazilian academic patents in the biotechnology and pharmaceutical technology fields by mining data contained in licensed and unlicensed patent documents. First, two databases were built: the first with the licensed patents from 8 universities that licensed the most patents in the pharmaceutical and biotechnology technological fields, and the second with the remaining characteristics of licensed and unlicensed patents. Next, three machine learning algorithms were assessed on the data, with the Random Forest algorithm achieving the best relative performance. Therefore, an RF model was developed to evaluate the licensing potential of these patents. At the end, the model performance was evaluated with metrics of accuracy, precision and recall. The result is an RF model for predicting the licensing potential of patents by determining the most influential variables.