Prediction of a Patent Property Using the Class Probability Output Network

Abstract

The maintenance period, the time frame begging from the registration to the expiration of a patent, is an important property used to evaluate the patent???s quality. To predict the maintenance period of a patent, a consistent classifier is desirable. The output of a classifier is usually determined by the value of a discriminant function and a decision is made based on this output which does not necessarily represent the posterior probability of the soft decision of classification. Thus, it is desirable that the output of a classifier be calibrated in such a way to include the posterior probability of class membership. For this purpose, the Class Probability Output Network (CPON) is devised. It is a new method of postprocessing for the probabilistic scaling of classifier???s output. To predict the maintenance period of a patent, SVM, KLR, perceptron and perceptron with CPON are used in simulation. The simulation results using the perceptron with CPON demonstrated a statistically meaningful performance improvement over that of SVM, KLR and perceptron.