面向智慧旅游的POI推荐

发布时间：2018-08-25 13:18

【摘要】：个性化的兴趣点推荐服务是一个重要而具有挑战性的工作,也是智慧旅游中的一项重要组成部分。以往的兴趣点推荐工作主要集中在本地用户的研究上,此类用户往往具有常规的活动范围,例如居住地区域和工作地区域,这些区域内的兴趣点具有更高的可能性被用户访问到,因此更加适合被推荐。然而,在许多实际的应用场景下例如旅游出行,目标用户往往是跨城市的旅行者。这类用户对于当地城市相当陌生,反应为用户在当地的访问数据非常稀疏,使得用户喜好建模变得非常苦难。因此如何有效地解决跨城市兴趣点问题就成为了一项巨大的挑战。本文试图从兴趣点维度出发,通过兴趣点之间的属性特征建立关联关系来提升跨城市兴趣点的推荐效果。对于空间上隔离的兴趣点,我们研究了人群分类特征(I),例如商务出行人群和家庭亲子人群。而针对本地兴趣点,我们则更加关注地理位置上的特性(N)。同时,语义上的特征(R)也被引入来解决空间隔离问题,并且使得推荐模型具备可解释性。融合上述三种兴趣点特征到隐空间的学习阶段,本文提出了一种新型的矩阵分解模型(INRMF)。本文主要贡献如下:·兴趣点特征挖掘对三种具有潜在价值的兴趣点特征进行建模和量化,分别为人群分类特征、地理位置特征和语义特征。通过评分层面的统计分析,发现了兴趣点在特征属性上呈现的关联关系,具有提升兴趣点推荐效果的积极作用。·旅游场景下兴趣点推荐基于矩阵分解技术,融合跨城市的人群分类特征和语义特征,以及本地的地理位置特征,提出了一种面向智慧旅游的跨城市兴趣点推荐买模型,有效解决数据稀疏问题,提升推荐效果。·语义特征优化应用短语级别的情感分析对评论文本进行挖掘,并通过词嵌入技术进行语义级别的优化,提升了词与词之间的关联关系,相应地提出两种基于矩阵分解的语义特征优化模型,提升了语义特征的利用效果。
[Abstract]:Personalized interest point recommendation service is an important and challenging task as well as an important component of smart tourism. Previous interest point recommendation work has focused on the research of local users, who often have a regular range of activities, such as residential and working areas, within these areas. Interest points are more likely to be accessed by users and therefore more suitable for recommendation. However, in many practical scenarios such as travel, the target user is often a cross-city traveler. Model becomes very difficult. Therefore, how to effectively solve the problem of cross-city interest points has become a huge challenge. This paper attempts to start from the dimension of interest points, through the attributes between interest points to establish a correlation relationship to enhance the recommendation effect of cross-city interest points. Class I features (e.g. business travelers and family parents and children). For local interests, we pay more attention to geographic features (N). Meanwhile, semantic features (R) are also introduced to solve the problem of spatial isolation and make the recommendation model explanatory. Learning from the above three interest features into hidden space is integrated. In this paper, we propose a new matrix decomposition model (INRMF). The main contributions of this paper are as follows: (1) Interest point feature mining is used to model and quantify three potential interest point features, namely, crowd classification features, geographical location features and semantic features. Interest point recommendation in tourism scenarios is based on matrix decomposition technology, which combines the characteristics of cross-city crowd classification and semantics, as well as the local geographical location features. A cross-city interest point recommendation and purchase model for intelligent tourism is proposed. Semantic feature optimization uses phrase-level emotional analysis to mine comment text, and optimizes the semantic level through word embedding technology to improve the relationship between words. Two optimization models of semantic feature based on matrix decomposition are proposed to improve the language performance. The utilization of semantic features.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【相似文献】