基于新颖性和多样性的旅游推荐模型研究

Research on tourism recommendation model based on novelty and diversity

人们在旅游活动中经常会利用推荐系统,比如推荐路线、推荐酒店等等,然而这种推荐多数是基于Top-N的热门项目推荐,经常导致游客得到一些信息量为0的"精准推荐"。针对传统的推荐算法过于强调推荐的精准度导致推荐列表的新颖性和多样性差的问题,将MMR技术应用在旅游推荐领域,同时加入用户-项目交互因子,提出一种基于发现的用户项目关系推荐模型,并在真实的数据集上进行测试,通过实验结果,和传统的KNN以及改进前的基于MMR经典算法对比,有效提高了推荐列表的新颖性和多样性。在旅游推荐这种新颖性较高的应用领域,该算法相对于传统的推荐算法具有较大的优势。

Recommendation system is widely used in people＇s travel activity, for example, recommendation of travel route and hotel. However, such kind of recommendation is mostly based on Top-N hot item recommendation, which leads to the fact that tourists end up with so-called accurate recommendation without any detailed information. This paper aims at the problem caused by over emphasis on recommendation accuracy in traditional recommendation algorithm that results in little novelty and diversity. It puts MMR into the application of tourism recommendation, and adds in user-item interaction parameter, thus putting forward a new recommendation system based on relationship between the user and the item.This new system undertakes tests in real data collection and compared with traditional KNN and classic MMR-based algorithm, effectively enhances the novelty and diversity of recommendation list. In such tourism recommendation field which requires higher novelty, this algorithm outweighs, to a certain extent, the traditional one.