Construction of a Prediction Model for Pharmaceutical Patentability Using Nonlinear SVM

摘要

The Japanese Patent Act follows a first-to-file principle, so it is crucial that important patent applications must be filed earlier than those by other inventors. However, inventors will not be awarded a patent if the description of the invention in the application is insufficient. Regarding this problem, a previous study investigated use of logistic regression in a prediction model for patentability (probability of acquiring patent rights). However, that model used linear discrimination, so the discrimination accuracy was not high. To increase prediction accuracy, this study instead uses a nonlinear support vector machine (SVM) in the predictive model for patentability. Evaluation experiments using the SVM model show that the prediction accuracy of the SVM-based model is better than that of the model used in the previous research. These results suggested that a nonlinear SVM model is effective for constructing a prediction model for pharmaceutical patentability.