基于复杂网络的产品质量安全事件演化与预测研究

The relevant regulatory authorities need to solve the urgent problem which is to enhance the management level in the field of product quality safety and master the initiation, development and evolution of product quality safety incidents. In this study, theories and methods of complex network are applied to compensate the shortage of evolution research for product quality safety incidents on social media with Web Mining. This project takes product quality safety incidents as the research object, intends to make use of complex network, Web mining, support vector machine(SVM) and Markov chain forecasting theory and method, research topic expression method of product quality safety incidents based on vector space model(VSM), propose its topic detection and tracking model based on feature words and incremental SVM; analyze the basic characteristics of the product quality safety incidents, research its evolution model based on the small-world network and scale-free networks; construct the trend assessment indicator system of product quality safety incidents, build the trend forecasting model based on the gray theory and weighted Markov chain, establish intervention model via the entropy method. Finally, inspection and correction for theoretical model is presented through case study with typical product quality safety incidents. This study will enrich the theories and methods of complex network modeling, incremental SVM and Markov, promote their applications in the field of product quality safety, and provide the support for the management, monitoring and decision-making of the product quality and safety incidents.

提升产品质量安全领域的管理水平，掌握产品质量安全事件产生、发展和演化规律，已成为相关监管部门亟待解决的问题。为了弥补Web挖掘在社会化媒体上产品质量安全事件演化研究的不足，本研究将复杂网络的相关理论和方法应用于其中。 本项目以产品质量安全事件为研究对象，利用复杂网络、Web挖掘、支持向量机和Markov链预测等理论和方法，研究基于向量空间模型的产品质量安全事件话题表达方法，建立特征词和增量SVM的主题检测和跟踪模型；分析产品质量安全事件基本特征，研究基于小世界网络和无标度网络的演化模型；研究产品质量安全事件态势评估指标体系，构建灰色加权Markov链的趋势预测模型，并利用熵权法建立干预模型。最后，通过典型产品质量安全事件案例研究对理论模型给予检验和修正。本研究有助丰富复杂网络建模和Markov等理论和方法，推动其在产品质量安全领域内的应用，并为产品质量安全事件的管理、监控和决策提供支持。

项目以产品质量安全事件为研究对象，提出了基于特征词和增量SVM的产品质量安全事件主题检测与跟踪算法，构建了基于复杂网络的产品质量安全事件演化模型，研究了产品质量安全事件的趋势预测，并以“毒跑道”事件为例开展了典型产品质量安全事件案例研究。项目共发表科技论文21篇，均标注自然基金号，其中，SCI检索7篇，EI检索12篇。出版论著1部，申请国家发明专利1项，获得软件著作权4项。