基于多属性打分的酒店推荐算法分析-analysis of hotel recommendation algorithm based on multi-attribute scoring.docxVIP

哈尔滨工业大学管理学硕士学位论文 - II - Abstract With the rapid development of network and information technology, faced with a flood of information resources, people are often difficult to find information to meet their own interests, and very easy to catch in the predicament of information overload. Therefore, the recommendation system in e-commerce came into being. E-commerce personalized recommendation as a kind of intelligent information services, provides users with information and services exactly which are interested through analyzing the user’ personality, habits and preferences. E-commerce personalized recommendation system can not only meet the needs of the user for personalized products, but also can improve the competitiveness of enterprises, so they are been widely used. Based on the research accomplishment both domestic and overseas and analyzing rating characteristics of the hotel on the internet, the paper studied on the collaborative filtering algorithm and its application in hotel recommendation, mainly include: Respectively, from the point of view of hotel suppliers and users, we analyzed the importance of hotel e-commerce personalized recommendation services, and the concept and role of e-commerce personalized recommendation system, and the main recommendation technology. The single overall score are used in the traditional collaborative filtering recommendation algorithm, and do not consider the users ratings on more than one criteria of the items, thereby affecting the accuracy and cannot reflect the users’ personalized characteristics more accurately. In this paper, we use multi-criteria hotel ratings on the Ctrip, propose to improve the traditional collaborative filtering algorithm based on the hotel multi-dimensional attributes scoring. On the one hand, we propose the algorithm based on the similarity of the expansion and distance-based collaborative filtering through improvement of the similarity calculation in the traditional collaborative filtering alg