【摘要】 为提高快递运输的风险监测管控能力,降低因快递货品风险导致城市安全事件发生的可能性,本文基于语义挖掘方法将快递运输货品描述转化为风险的量化表征,为快递运输风险评价提供可量化的客观指标依据。基于网络大数据资源提供的法院判决书数据,将物品词条与判决结果相关联,通过隐狄利克雷分布模型挖掘物品风险主题,结合模糊均值聚类方法,实现对快递货品语义风险的量化表征与柔性划分。与传统方法中依赖检视人员查验既定违禁品清单后的主观判断方法不同,本文充分挖掘网络文本数据中的可迁移知识,并应用于种类繁多的快递运输货品,有效避免人工评价造成的漏检、错检情况。研究结果表明,本文方法具有较高的准确率与较低的误报率,获得的风险评价值不再是0或1的是非判断,有利于开展多样化、针对性的风险预警及应对措施。更多还原

【Abstract】 To improve the risk monitoring of express transportation and reduce the possibility of urban security incidents caused by risky express goods, this paper transforms the description of express delivery goods into the quantitative representation of risks based on the semantic mining method. The study provides a quantifiable and objective index basis for risk evaluation of express transportation. Based on the court verdict data provided by the Web big data resources, this study correlates the goods descriptions with the verdicts and explores the goods risk topics through the hidden Dirichlet distribution model. The fuzzy C-means clustering method is also incorporated to realize the quantitative representation and flexible division of the semantic risk of express delivery goods. Different from the traditional method which relies on the subjective judgment of the inspector after checking the fixed contraband list, the proposed method fully excavates the transferable knowledge from the Web text data and applies it to different types of express delivery goods, which can effectively avoid missing inspections and errors caused by manual evaluation. The results show high accuracy rate and low false alarm rate, and the obtained risk evaluation value is no longer“0”or“1”representing the judgment of“yes”or“no”. The proposed method helps to make diversified and focused early warning and response measures to avoid risks in express deliveries.更多还原