基于Voting集成学习的颠覆性专利识别模型构建及应用——以手机通讯领域为例

Construction and Application of Disruptive Patent Identification Model Based on the Voting Ensemble Learning——Taking the Field of Mobile Communication as an Example

摘要:

颠覆性技术是新质生产力形成和发展的核心驱动力与关键要素，颠覆性专利则是颠覆性技术的主要表现形式，做好颠覆性专利的早期识别与预测，有助于前瞻性地优化专利布局策略、运营方案及资源配置。本研究旨在构建一个高效、准确的颠覆性专利自动识别模型，着力解决从海量专利数据中发现颠覆性专利的难题。以手机通信领域为例，从相关专利文献中提取15个特征项，构建基于Voting集成学习的颠覆性专利识别模型，将其应用于潜在颠覆性专利识别及专利颠覆性指数测度，并对模型有效性进行检验。该模型的准确率76.53%，召回率76.63%，AUC值0.85，F1值75.07%，高于单一算法模型，证实基于集成学习的识别和预测效果优于单一算法，且指标可获得性强、模型操作简便，适用于面向专利大数据的颠覆性技术识别与预测场景。

Abstract:

Disruptive technology is the core driving force and key element for the formation and development of new quality productive forces, and disruptive patents are the main manifestation of disruptive technology. Early identification and prediction of disruptive patents can help optimize patent layout strategies, operational plans and resource allocation with more foresight. This study aims to construct an efficient and accurate automatic identification model for disruptive patents, focusing on solving the challenge of discovering disruptive patents from massive patent data. Taking the field of mobile communication as an example, this study extracted 15 feature items from patent literatures and constructed a disruptive patent identification model based on the Voting ensemble learning. The model was applied to identify disruptive patents and measure their disruptive index, as well as the effectiveness of the model was tested. The accuracy of the model is 76.53%, the recall rate is 76.63%, the AUC value is 0.85, and the F1 value is 75.07%, which is higher than that of a single algorithm model. It approved that the recognition and prediction performance based on ensemble learning is better than that of a single algorithm, and the indicators are highly obtainable and the operation is simple. It is suitable for big data patent prediction scenarios.