首页|基于多模态数据融合与机器学习的高价值专利早期识别方法

基于多模态数据融合与机器学习的高价值专利早期识别方法

扫码查看
原文链接
  • 万方数据
  • 维普
[研究目的]如何构建有效的高价值专利识别方法,从海量的专利中识别出高价值专利,对于高价值专利的培育、运用、保护和管理具有重要意义.[研究方法]提出一种基于多模态数据融合与机器学习的高价值专利早期识别方法.首先,利用多模态融合方法对专利指标数据和文本数据两种模态数据进行融合;其次,利用基于机器学习的生存预测算法对历史已知的高价值专利进行分析,并获取高价值专利融合后的特征与其"高价值"之间的关联模式;最后,利用获取到的关联模式,对新公开专利的"价值"进行预判,从而实现对高价值专利的早期识别.[研究结论]以人工智能领域专利为例进行实证研究,验证了该方法的可行性和有效性,为高价值专利早期识别提供了新的研究方法.
Research on Early Identification Method of High Value Patents Based on Multi-Modal Data Fusion and Machine Learning
[Research purpose]How to build an effective method for identifying high-value patents from a massive number of patents is of great significance for the cultivation,application,protection,and management of high-value patents.[Research method]We propose an early identification method of high-value patents based on multi-modal data fusion and machine learning.Firstly,uses a multi-modal fu-sion method to fuse two modal data:patent indicator data and text data;secondly,uses a survival prediction algorithm based on machine learning to analyze high-value patents known in history in a certain field,and obtain the correlation model between the fused features of high-value patents and their"high value";finally,uses the obtained correlation model to predict the"value"of the patent,thereby achieving early identification of high-value patents in this field.[Research conclusion]An empirical study is conducted taking patents in the field of artificial intelligence as an example,the feasibility and effectiveness of the proposed method is validated by the case results,providing a new research method for early identification of high-value patents.

multimodal data fusionhigh value patentspatent evaluationsurvival prediction algorithmmachine learningearly identifi-cationartificial intelligence

李欣、李悦、冯野、高宁

展开 >

北京工业大学经济与管理学院 北京 100124

多模态数据融合 高价值专利 专利评估 生存预测算法 机器学习 早期识别 人工智能

国家自然科学基金面上项目国家电网科技项目

721740171400-202255271A-2-0-XG

2024

10.3969/j.issn.1002-1965.2024.06.019

情报杂志
陕西省科学技术信息研究所

情报杂志

CSTPCDCSSCICHSSCD北大核心
影响因子：1.502
ISSN：1002-1965
年,卷(期)：2024.43(6)