自2000年以来，中国进入了一个灾害的新时期，气象灾害事件频繁发生，同时现有的应急预案数量越来越庞大，质量又参差不齐。为了快速准确地为某具体气象灾害事件提供合适的、参考价值高的应急预案，提高应急预案编制工作的效率和质量，推荐技术被应用到了应急预案领域。然而，现有的应急预案推荐研究多是针对人做推荐，考虑的是用户的个人喜好、需求问题，而实际上我们往往需要关注的是针对一个气象灾害事件推荐参考价值高的应急预案；此外，传统的推荐算法存在冷启动、评分矩阵稀疏、忽略语义信息等问题。为了解决以上这些问题，本文首次进行了对气象灾害事件产生应急预案推荐的算法的研究，并且引入了本体技术，考虑了语义知识等，最终提出了一个语义可知的气象灾害应急预案推荐算法。本文的主要工作和创新点体现在： （1） 将推荐系统的研究首次应用到气象灾害应急预案领域，并且切合实际状况，对气象灾害事件而非用户进行准确的应急预案推荐，缓解了气象灾害应急预案数量日渐庞大、质量参差不齐而带来的寻找有参考价值的应急预案耗时的问题，促进应急管理的进步。 （2） 建立了统一的气象灾害事件应急情景本体模型。本文深入研究气象灾害事件的特征，考虑了气象灾害事件的应急情景信息，包括致灾因子、自然环境情景和社会环境情景，通过 Protégé 建立了气象灾害事件应急情景语义本体，并且复用了地区本体，统一表示各类气象灾害事件。 （3） 通过事件之间的应急情景信息和地理信息计算事件之间的相似度，寻找目标事件的邻居，进而产生推荐，克服了冷启动问题。 （4） 提出了语义可知的气象灾害应急预案推荐算法。我们的算法是以基于用户的协同过滤算法为基础进行设计的。在寻找邻居时，一方面我们引入了事件语义相似度这一概念，改进事件相似度计算公式，考虑了语义知识，另一方面，我们提出了新的权重计算方法FMWC来灵活地调整语义特征权重，尽量避免了假邻居的产生和真邻居的缺失。 （5） 使用jena推理机模拟了我们的推荐算法，实验表明本文提出的算法不仅解决了新事件的冷启动问题，而且有较高的推荐准确率。

Since 2000, China has entered a new era of disaster, Meteorological disaster events occur frequently, and the number of existing emergency preplans is increasing and the quality is uneven. In order to provide a suitable preplan with high reference value for a specific meteorological disaster event quickly and accurately, improving the efficiency and quality of emergency preplan preparation, recommendation technology has been applied to the area of emergency preplan. However, most of the current research on emergency preplan recommendation is to make recommendations for people, considering the users' personal preferences and needs, in fact, what we often need to pay attention to is to how to make recommendations for events; in addition, traditional recommendation algorithms always have the problems of cold start, sparse scoring matrix, and ignoring semantic knowledge. In order to solve these problems, this paper first carried out the research on the algorithm for the recommendation of emergency preplans for meteorological disaster events, introduced the ontology technology, and considered the semantic knowledge, a method called A Semantic Aware Emergency Preplan Collaborative Filtering Recommendation Method for Meteorological Disaster Events is put forward. The main work and innovation of this article are as follows: (1) The application of recommendation system is first applied to the field of meteorological disaster emergency preplan, and in accordance with the actual situation, recommending emergency preplans for meteorological disasters rather than users accurately. It alleviated the time-consuming problem of finding suitable emergency preplans caused by the increasing number and the uneven quality of emergency preplans, promoting the progress of emergency management. (2) A unified emergency scenario ontology model for meteorological disasters has been established. We studied characteristics of meteorological disaster events, considering emergency scenario information of meteorological disaster events, including disaster causing factors, natural environment scenarios and social environment scenarios. Then we established semantic ontology of emergency scenarios for meteorological disasters by Protégé, and reused the regional Ontology, representing all kinds of meteorological disasters in a unified way. (3) The similarity between events is calculated through emergency scenario information and geographic information, to find the neighbor of the target event, and then produce the recommendation, which overcomes the problem of cold start. (4) A Semantic Aware Emergency Preplan Collaborative Filtering Recommendation Method for Meteorological Disaster Events is proposed. Our algorithm is designed on the basis of user-based collaborative filtering algorithm. When looking for a neighbor, on the one hand, the concept of event semantic similarity is introduced to improve the calculation formula of event similarity and semantic knowledge is considered; on the other hand, a new weight calculation method FMWC is proposed to adjust the semantic feature weight flexibly, which can avoid the production of false neighbors and the missing of real neighbors. (5) Using Jena reasoning machine to simulate our recommendation algorithm, the result shows the algorithm proposed in this paper not only solves the cold start problem of new events and has high recommendation accuracy.