基于样本均衡技术的专利侵权预警研究
Research of Patent Infringement Early Warning Based on Sample Equilibrium Technology

知识产权的保护影响科研人员的创造力和企业的研发动力，是创新的重要驱动力和法律保障，专利侵权预警是知识产权保护的重要组成，而大量的专利侵权诉讼案件数据为挖掘、分析专利侵权行为模式，发现专利侵权的风险提供了基础。本文利用专利相关的特征信息，基于样本均衡技术构建了专利预警模型，对比分析了随机森林、贝叶斯网络、神经网络、决策树模型、逻辑回归模型和SVM算法的性能。实验结果表明，随机森林模型在样本均衡后可取得更好的预警效果，能够更好地发现公司与专利之间的侵权诉讼关系，从而有效地实现专利侵权预警的功能。
The protection of intellectual property rights affects the creativity of researchers and development power of enterprises. It is an important driving force and legal guarantee for innovation. Patent infringement warning is an important component of intellectual property protection. A large number of patent infringement lawsuit data are used to analyze the model of patent infringement and to discover the risks of patent infringement. Based on the patent-related feature information, this paper constructs a patent early warning model based on sample equalization technology, and compares the performance of random forest, Bayesian network, neural network, decision tree model, logistic regression model and Support Vector Machine (SVM) algorithm. The experimental results show that the random forest model can obtain better early warning effect after sample equilibrium, and can better discover the infringement litigation relationship between the company and the patent, thus effectively realizing the function of patent infringement warning.