基于协同过滤法的旅游目的地推荐系统研究

发布时间：2018-05-10 06:02

本文选题：协同过滤法 + 推荐技术　；参考：《中国海洋大学》2013年硕士论文

【摘要】：伴随着信息技术的不断发展，旅游者的消费模式发生了重大转变，对信息的依赖程度不断增强，对网络资源的应用也在不断扩展，对信息推荐技术地应用和推广便是其中的重要代表。信息推荐技术帮助用户筛选过滤不必要信息，将用户需求的最有效的信息传达给用户，降低用户搜集、整理、分析资料的时间和资金成本。信息推荐技术主要包括协同过滤法和内容分析法，该文在研究过程中，将协同过滤法作为研究的主线引入到旅游领域，同时针对目的旅游资源利用方面借助了内容分析法的重要思想，形成了协同过滤法为主线，内容分析法为重要辅助的有机结合模式。在该文中，加强了对推荐技术、协同过滤法、内容分析法、DMS系统、旅游目的地营销相关理论内容的研究介绍和区分，理顺了文章的理论基础和各理论间的关联性。在此基础上，针对旅游目的地推荐系统的基本要素和运算过程做了进一步研究。针对推荐系统基本要素探讨，主要对以下三方面进行研究：目的地特征向量、旅游者经验向量、旅游者特征向量。目的地特征向量方面主要包括：（1）目的地信息输入主体的研究，最终确定了旅游目的地管理部门作为信息输入主体，并确定了采用了国家标准法对目的地旅游资源进行划分；（2）目的地输入表格模版研究，输入内容包括：目的地所在区域、目的地名称、星级评定、电话号码、旅游管理部门、网址、旅游特征、目的地开发时间。较为具体形象地表明了目的地特征，提升了旅游者对目的地信息的把握能力。旅游者经验向量方面：本目的地推荐系统重视旅游者的旅游经历，通过对与具有相同喜好的旅游者进行相似度计算，根据相似度较高的旅游者经历，向目标旅游者推荐新的旅游目的地，实现目的地推荐。在设计输入表格模板过程中，考虑到存在以下三种情况：（1）已经输入过旅游信息的旅游者；（2）新加入的没输入信息的旅游者；（3）新目的地项目的加入。针对这三种情况分别设计了不同的输入模板以适应旅游者不同的需要。旅游者特征向量方面：本研究中设定旅游者t的特征向量为Ft，通过旅游者经验矩阵H和目的地特征矩阵E自动计算得到。旅游者访问的目的地越多，目的地旅游特征越集中，便越能准确地反映出旅游者的个人喜好。通过对了推荐系统基本要素探讨，最后针对旅游目的地推荐的形成的具体运算方法进行了论述，提出了基于项目分类的协同过滤改进算法该推荐方法。首先利用项目分类信息采用聚类技术为类内未评分项目预测评分值，弥补协同过滤法存在的数据稀疏性问题。然后通过计算类内用户间的相似度得到目标用户的最近邻居，最后进行推荐。在本文的最后，针对研究过程中出现的问题，为推进系统未来的应用研究，提出以下建议：（1）通过在实践中不断丰富完善协同过滤推荐方法，保证该系统能及时准确地反映旅游者兴趣的变化时间和趋势，准确把握不同时间阶段旅游者兴趣特征，通过不断阶段的聚类分析，推进对旅游者相似度的把握。（2）推进对反映旅游者兴趣的指标建设。可以通过设计体现旅游兴趣反应的指标，由旅游者自行输入信息，，通过软件分析，事先注意到旅游者不同阶段的特征，及时把握旅游者兴趣变化。根据不同阶段的兴趣特征，辅助协同过滤法推进旅游者兴趣聚类，进而提升向不同阶段旅游者推荐目的地的准确度和满意度。
[Abstract]:With the continuous development of information technology, the consumption pattern of tourists has changed greatly, the dependence on information is increasing, the application of network resources is expanding, and the application and popularization of information recommendation technology is an important representative. The most effective information of the demand is conveyed to the user, reducing the user collection, sorting, and analyzing the time and cost of the data. The information recommendation technology mainly includes collaborative filtering and content analysis. In the study, the collaborative filtering method is introduced as the main line of research into the tourism field, and the use of tourism resources is also aimed at the purpose of the study. With the help of the important idea of content analysis, the collaborative filtering method is the main thread, and content analysis is an important supplementary mode.
In this article, we have strengthened the introduction and distinction of the relevant theoretical contents of the recommendation technology, collaborative filtering, content analysis, DMS system and tourism destination marketing, and rationalized the theoretical basis of the article and the relevance between the various theories. On this basis, the basic elements and operation process of the tourism destination recommendation system were further studied. In view of the basic elements of recommendation system, the following three aspects are studied: destination feature vector, tourist experience vector, tourist characteristic vector.
The feature vector of destination mainly includes: (1) the research of destination information input subject, and finally determines the tourism destination management department as the information input subject, and determines the use of national standard method to divide the destination tourist resources; (2) the destination input form template study, including the destination location, is the destination Area, destination name, star rating, telephone number, tourism management department, website, tourist characteristics, destination development time. The specific image shows the destination features and enhance the ability of tourists to grasp the information of the destination.
In the aspect of tourists' experience vector: this goal recommends the system to pay attention to tourists' travel experience. By calculating the similarity with the tourists with the same preference, according to the experience of the tourists with higher similarity, the new tourist destination is recommended to the target tourists and the destination recommendation is realized. The following three situations are considered: (1) tourists who have already entered the tourist information; (2) newly joined tourists without input information; (3) the addition of new destination projects. Different input templates are designed for these three situations to adapt to the different needs of tourists.
In the aspect of tourists' characteristic vector, the characteristic vector of tourist t is set as Ft in this study. The tourist experience matrix H and the destination feature matrix E are automatically calculated. The more destination the tourists visit, the more concentrated the destination tourism features, the more accurate the tourist's personal preferences can be reflected.
Based on the discussion of the basic elements of the recommendation system, the concrete operation methods of the tourism destination recommendation are discussed, and the proposed method of collaborative filtering improved algorithm based on project classification is proposed. First, the clustering technology is used to make up the evaluation of the non scoring items in the class by using cluster technology to make up for the collaborative filtering. The problem of data sparsity exists in the method. Then, the nearest neighbor of the target user is obtained by calculating the similarity between the users in the class. Finally, in this paper, the following suggestions are put forward in order to promote the future application research of the system in order to solve the problems in the process of the study.
(1) through constantly enriching and perfecting the collaborative filtering recommendation method in practice, the system can ensure that the system can reflect the changing time and trend of tourists' interest in a timely and accurate way, accurately grasp the characteristics of tourists' interest in different time stages, and promote the similarity of tourists through continuous stage clustering analysis.
(2) to promote the construction of the indicators that reflect the interest of the tourists. Through the design of the indicators that reflect the response of the tourist interest, the tourists can input the information by themselves. Through the software analysis, we can notice the characteristics of the tourists in different stages in advance and grasp the changes of tourists' interest in a timely manner. According to the characteristics of different stages of interest, the cooperative filtering method is used to promote tourism. The clustering of interest will further enhance the accuracy and satisfaction of destination recommendation for tourists at different stages.

【学位授予单位】：中国海洋大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F590.3

【参考文献】