基于空间数据挖掘的热门景点及线路推荐研究

发布时间：2018-03-04 23:16

本文选题：空间数据挖掘　切入点：地理标签照片　出处：《华中师范大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着科技的发展和人们生活水平的提高,旅行已经成为越来越多的人生活中至关重要的一环,虽然现在有关旅行的应用系统很多,但真正能够满足用户实际需求的却很少,例如通过搜索引擎或者旅行网站搜索的结果往往是旅行公司的商业推广或者旅游景点的过度宣传。如何让游客在一个陌生的城市中快捷方便地找到热门旅游景点并规划好游玩路线,这是智慧旅游时代自助旅游者的迫切需求。本文总结了国内外关于空间数据挖掘和旅游推荐的相关技术和方法,并在此基础上提出了一种基于空间数据挖掘的热门景点和线路推荐方法。由于大部分旅行者都有拍照的习惯并且喜欢在社交媒体上分享这些照片,本文首先从大型图片分享网站Flickr上挖掘了大量带有空间地理信息和文本Tag的Geo图片,通过 P-DBSCAN(A Parallel of Density-Based Spatial Clustering of Applications with Noise, 一种并行的基于密度的聚类算法)聚类算法对这些用户上传的携带空间地理信息的图片进行空间聚类得到图片所在的景点集,然后根据文本Tag将这些聚类得到的景点与真实的景点进行对应,并构建景点数据库。然后,计算游客对不同景点的游览次数来找出热门景点,根据游客的旅行历史和背景信息等利用协同过滤算法计算用户相似度,并以此为依据进行个性化景点推荐,在推荐的同时考虑天气和时间等因素。最后,对游客的旅行历史进行分析可以提取出游客的旅行轨迹,使用序列模式挖掘算法分析出热门旅行路线,再根据游客的旅行历史和当前的搜索文本来进行个性化的线路推荐。本文在聚类的过程中采用了 P-DBSCAN算法,通过实验证明P-DBSCAN算法在密度分布不均的情况下聚类效果明显优于DBSCAN算法。而在推荐过程中引用了 Apache Mahout提供的一个协同过滤算法的推荐引擎Taste,并在Android设备上展示最终推荐结果。
[Abstract]:With the development of science and technology and the improvement of people's living standard, travel has become a vital part of life more and more people, although the application system about traveling a lot, but be able to meet the actual needs of users rarely, for example through search engines or travel site search is often the result of excessive promotional or commercial Travel Company tourist attractions. How to let visitors in a strange city in fast and convenient to find popular tourist attractions and planning to visit route, this is the urgent demand of wisdom tourism era self-help tourists. This paper summarizes the related technologies of spatial data mining and tourism recommendation and methods at home and abroad, and put forward a based on spatial data mining for hot spots and lines. The recommended method of most of the travelers have the habit of taking pictures and in love Social media to share these photos, this paper from the large photo sharing website Flickr mining with a lot of spatial geographic information and text Tag Geo images, through P-DBSCAN (A Parallel of Density-Based Spatial Clustering of Applications with Noise, a parallel clustering algorithm based on density clustering algorithm) for these users to upload carry geographical space the picture information spatial clustering images obtained by the attractions set, and then the clustering obtained spots corresponding to the real attractions according to the text Tag, and construct the scene points according to the library. Then, the calculation of the number of tourists to visit different scenic spots to find hot spots, calculated according to the similarity of users travel history and background information by tourists the collaborative filtering algorithm, and as a basis for personalized recommendation, and consider the weather in the recommended at the same time Among other factors. Finally, analysis can extract the travel trajectory of tourists on the tourist travel history, mining algorithm analysis hot travel route using the sequence pattern, according to the tourists travel history and the current search text to personalize the lines recommended. This paper uses the P-DBSCAN algorithm in the clustering process, proved by experiments P-DBSCAN clustering algorithm in the case of uneven density distribution is better than the DBSCAN algorithm. In the process of recommendation refers to a collaborative filtering recommendation engine Taste Apache provided by Mahout, and show the final recommendation on the Android devices.

【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13;TP391.3

【参考文献】