云南旅游推荐系统信息提取与推荐算法研究

发布时间：2019-06-21 04:31

【摘要】：近些年,由于人们生活水平提高与日常工作压力增大,随着旅游业越来越热门,大众的需求刺激,大量的旅游网络服务公司在市场上涌现。通过观察发现,现今旅游网站上对用户的推荐仍以传统推荐为主,向用户推荐热度较高的景点或做基于协同过滤的推荐,但由于云南省全省温度差异较大,并且由于景点宣传力度不同,大部分风景优美的景点并不为人所知。这些推荐显然都不太符合云南省旅游业的实际情况并且无法满足用户的需求。用户通过网络旅游平台定制门票或出行计划,并发表评价或书写自己的旅游文记,同时也有人们在出行前会浏览这些信息以供参考,但是大量的信息浏览会给人们造成困扰,并且大量的文本信息得不到合理利用也会造成资源浪费。由此,本文致力于挖掘这些文本信息并结合云南的季节因素、景点的热度因素对用户做出关于云南旅游景点的个性化推荐系统。文中提出了两种针对云南旅游的推荐算法:第一种是系统冷启动状态下,提出的基于景点内容的推荐算法。该方法利用每个景点的评价文本集合进行LDA主题模型的文档主题自动提取,进而对所有景点生成的主题-词空间向量矩阵进行k-means聚类,并为得到的各个类别进行人工标注类别名称。由于云南旅游的特殊性,本文通过景点评价信息的数量与发表时间,设计了季节权重与热度权重。通过用户进入系统,系统主动问询,用户手动选择类别名称的方式,获取用户的兴趣点,在此基础上,系统根据用户兴趣类别结合实时季节权重与景点热度权重为用户推荐当前季节最适合游玩并符合用户兴趣的景点,达到了满足用户兴趣、并全面宣传云南旅游景点的效果;另外一种算法是系统存在单个用户历史行为记录下,提出的基于旅游文记信息的景点关联规则的推荐算法。通过对用户与系统的交互行为数据的分析,得到用户对景点的兴趣度和认同度,在此基础上,利用旅游文记中的景点名词形成各个用户到达过的景点集合,运用关联规则算法挖掘出景点之间隐含的关联,作为系统推荐景点的根据,满足了不同群体的需求,实现了个性化的服务。该推荐系统不仅能够取得大众群体的共性需求,也能够识别出隐含的适用于个性化需求的规则,并且在推荐的实时性与准确性上都有较好的效果,可以满足用户对景点推荐的需求。
[Abstract]:In recent years, due to the improvement of people's living standards and the increasing pressure of daily work, with the increasing popularity of tourism and the stimulation of public demand, a large number of tourism network service companies have emerged in the market. Through observation, it is found that the recommendation to users on the tourism website is still based on the traditional recommendation, and the recommendation based on collaborative filtering is recommended to the users. However, due to the large temperature difference in Yunnan Province and the different publicity of scenic spots, most of the scenic spots with beautiful scenery are not known. These recommendations are obviously not in line with the actual situation of tourism in Yunnan Province and can not meet the needs of users. Users customize tickets or travel plans through the online tourism platform, and publish evaluation or write their own travel notes. At the same time, some people will browse these information for reference before traveling, but a large number of information browsing will cause trouble to people, and a large number of text information can not be used reasonably will also cause waste of resources. Therefore, this paper focuses on mining these text information and combining the seasonal factors of Yunnan and the heat factors of scenic spots to make a personalized recommendation system for users about Yunnan tourist attractions. In this paper, two recommendation algorithms for Yunnan tourism are proposed: the first is the recommendation algorithm based on scenic spot content in the cold start state of the system. This method uses the evaluation text set of each scenic spot to extract the document theme of LDA topic model automatically, and then carries on the k-means clustering to the topic-word space vector matrix generated by all scenic spots, and carries on the manual label category name for each category. Because of the particularity of Yunnan tourism, this paper designs the seasonal weight and heat weight through the quantity and publication time of scenic spot evaluation information. Through the user entering the system, the system takes the initiative to inquire, the user manually selects the category name way, obtains the user's interest point, on this basis, according to the user interest category unifies the real-time seasonal weight and the scenic spot heat weight to recommend the scenic spot which is most suitable for the current season and conforms to the user's interest, achieves the satisfaction user interest, and comprehensively propagandizes the Yunnan tourist attraction effect. Another algorithm is a recommendation algorithm of scenic spot association rules based on tourism record information, which is based on the historical behavior record of a single user in the system. Through the analysis of the interactive behavior data between the user and the system, the interest degree and identity degree of the user to the scenic spot are obtained. on this basis, the scenic spot collection reached by each user is formed by using the scenic spot nouns in the tourist records, and the implicit association between the scenic spots is excavated by using the association rule algorithm, which is used as the basis of recommending scenic spots in the system, which meets the needs of different groups and realizes individualized service. The recommendation system can not only obtain the common needs of the mass group, but also identify the implicit rules suitable for personalized requirements, and has a good effect on the real-time and accuracy of the recommendation, which can meet the needs of users for scenic spot recommendation.
【学位授予单位】：云南财经大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F592.7

【参考文献】