Abstract: Industry 4.0, represented by disruptive technologies such as virtual reality, cloud computing, and the Internet of Things, has led to the reshuffling of global traditional industries. It brings industrial transformation in various countries, reinvention and convergence of industries, forming a new industrial standards, industrial patterns and business models. The huge demand of national economic development for disruptive technologies makes it an important prerequisite to seize the initiative of innovation to accurately identify the field of disruptive technologies, accurately forecast the direction of disruptive technologies, and correctly guide the development of disruptive technologies. Based on the characteristics of cross-border convergence and diffusion of disruptive technologies, this paper proposes a dynamic prediction model architecture of disruptive technologies based on patent data from the perspective of technology convergence. Virtual reality, a representative disruptive technology, is selected and serves as the empirical object of this paper. The life cycle of this technology is creatively fitted by constructing a patent co-occurrence network and analyzing the cumulative number of patent co-occurrence, with 2010 identified as the dynamic starting point for the current prediction phase. To address data disequilibrium, a dynamic prediction model is developed using link similarity indices and machine learning classification algorithms. A comparison of various prediction models shows that the global similarity index outperforms the local similarity index in forecasting subversive technologies, and the random forest algorithm emerges as the most effective classifier. The prediction results suggest that the most promising areas for technological convergence with virtual reality include near-eye display devices, digital data processing technologies, and digital video transmission systems. By scientifically designing the prediction process for disruptive technologies, this paper not only effectively improves forecasting accuracy, but also provides theoretical foundation and methodological reference for identifying critical timing and selecting appropriate prediction models for the development of disruptive technologies.

Key words: disruptive technology, technology convergence, technology life cycle, link prediction, machine learning

摘要： 虚拟现实、云计算、物联网等为代表的颠覆性技术正推动工业4.0的发展，引发全球传统产业深度变革,促使各国工业体系的融合，并逐步形成新的工业标准、产业格局和商业模式。由于国家经济发展对颠覆性技术的需求日益增长，准确识别颠覆性技术领域、精确预测其发展方向并合理引导其发展，已成为抢占创新主导权的关键前提。基于颠覆性技术跨界融合与扩散的特点,从技术融合的视角出发，提出一种基于专利数据的颠覆性技术动态预测模型框架。首先，以典型颠覆性技术———虚拟现实技术为实证对象，创造性地构建专利共现网络，并利用专利共现累计数量拟合技术生命周期曲线，动态识别出2010年为预测虚拟现实技术的起始时间节点。其次，针对数据的非均衡特征，采用链路相似性指标结合机器学习分类算法，构建动态预测模型。最后，通过对比不同的预测模型，研究发现：颠覆性技术预测中，全局相似性指标优于多数局部相似性指标，而随机森林算法是最佳分类器。基于此，研究进一步预测虚拟现实技术最具有发展潜力的融合领域为近眼显示设备、电数字数据处理技术和数字视频传输系统。本研究通过科学设计颠覆性技术的预测流程，显著提升了预测准确性，不仅为识别颠覆性技术发展的关键时间节点提供了理论依据，也为预测模型的选择与优化提供了方法参考。

关键词: 颠覆性技术, 技术融合, 技术生命周期, 链路预测, 机器学习