基于混合协同过滤的旅游攻略推荐算法研究

发布时间：2018-02-02 06:48

本文关键词： 推荐系统推荐算法旅游攻略覆盖度协同过滤　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：大数据的逐步发展使人类迈入了信息过载的时代,用户怎么样才能在纷繁复杂的信息中择优选出他们所需的信息,信息又如何有效的呈现在用户面前呢,这无疑给计算机工作者带来了新的挑战。为解决此问题,推荐系统随之而生。它是解决信息过载问题的重要方法,并且与搜索引擎相比,它可以为其私人订制,满足每一个用户的独特需求。用户在使用网站时的行为数据会被记录下来,推荐系统通过分析这些数据获取每一个用户兴趣所在,从而投其所好,为每一个用户寻得他们需要的信息。如今,推荐系统被应用的领域越来越广泛,其商用价值也越来越大,也得到了学术界越来越多的关注和研讨,不仅在理论上有很大提升,更在实践方面有质的飞跃,逐步形成了一门独立的学科。在新的时代,推荐系统也面临一系列挑战。论文首先论述了推荐系统的研究背景及意义,并介绍其在国内外的研究现状。之后详细讲解并比较了两个基本算法,首先介绍的一种协同过滤推荐算法是基于用户的,随后本文又通过案例分析介绍了另一种协同过滤推荐算法,该算法是基于物品的[1]。紧接着在下一章中介绍了用于评价推荐系统好坏的评测指标。最后,本文针对旅游攻略提出了基于混合协同过滤的旅游攻略推荐算法。旅游攻略是旅行市场的一款产品,由用户创建用户阅览,优质攻略平均每篇文字数万,覆盖多个城市,这些一般是相邻的城市,如果是国外,一般同一个国家也有少数跨同一个洲。每篇攻略都包含标题、内容、旅游城市、游玩景点,其中的旅游城市是攻略推荐中最核心的三个因素之一。本文研究的基于混合协同过滤的旅游攻略推荐系统是针对旅游市场设计的个性化推荐系统,结合两种基本算法,经过不断分析与实验,通过优化推荐度来对算法加以改进,以提高覆盖度。主要改进的地方有两点,第一点是对用户和物品的兴趣度加以平衡,第二点是加入城市热度和攻略热度的参数,并对其加以惩罚,目的是使热度高的城市或攻略推荐度低一些,使那些冷门的城市或攻略能更容易的被推荐出去。本文对基于旅游用户、旅游攻略和旅游城市推荐的推荐度分别进行了计算,然后平衡三个推荐度得到最终的推荐度。综合考虑了旅游攻略的数据特点和推荐过程的特殊性,极大程度增加推荐攻略的覆盖度并解决“长尾效应”的推荐问题,提出了一个针对旅游攻略的极具特色的推荐系统。
[Abstract]:Big data's gradual development makes people enter the era of information overload, how can users choose the information they need in the complicated information, and how to effectively present the information in front of the users? This undoubtedly brings a new challenge to computer workers. In order to solve this problem, recommendation system comes into being. It is an important method to solve the problem of information overload, and compared with search engine, it can be customized for private users. To meet the unique needs of each user. The user behavior data in the use of the site will be recorded, the recommendation system through the analysis of these data to obtain each user interest, so as to take advantage of it. Nowadays, recommendation system has been applied more and more widely, and its commercial value has become more and more great, and it has also been paid more and more attention and research in academia. Not only in theory there is a great improvement, but also in practice there is a qualitative leap, gradually formed an independent discipline. In the new era. Recommendation system also faces a series of challenges. Firstly, this paper discusses the research background and significance of recommendation system, and introduces its research status at home and abroad. Then, two basic algorithms are explained and compared in detail. This paper first introduces a collaborative filtering recommendation algorithm based on user, and then introduces another collaborative filtering recommendation algorithm by case study, which is based on articles. [1. Then in the next chapter, the evaluation index used to evaluate the quality of recommendation system is introduced. Finally. This paper puts forward a recommendation algorithm of tourism strategy based on mixed collaborative filtering. Tourism strategy is a product of the travel market. Users create users to read, and the average quality of each article is tens of thousands of words. Covers many cities, these are generally adjacent cities, if abroad, generally the same country has a few across the same continent. Each strategy contains the title, content, tourist cities, tourist attractions. The tourism city is one of the three core factors in the strategy recommendation. The tourism strategy recommendation system based on mixed collaborative filtering is a personalized recommendation system designed for the tourism market. Combined with two basic algorithms, through continuous analysis and experiments, the algorithm is improved by optimizing the recommendation degree to improve the coverage. There are two main improvements. The first is to balance the interest of the user and the object, and the second is to add and punish the parameters of urban heat and strategy heat, with the aim of making cities with high heat or strategic recommendations lower. So that those unpopular cities or strategies can be more easily recommended out. This paper calculates the recommended degree based on tourism users, tourism strategy and tourism city recommendation. Then the final recommendation degree is obtained by balancing the three recommended degrees. The characteristics of the data of tourism strategy and the particularity of the recommendation process are considered synthetically. By greatly increasing the coverage of the recommended strategy and solving the problem of "long tail effect", a special recommendation system for tourism strategy is proposed.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【相似文献】