基于社交媒体地理大数据的可感知情境的个性化旅游推荐研究

In the era of the Internet, travel recommendation system is an important means to help the tourist filter complicated information and make a sensible decision. It has become one of the research hotspot in tourism. Besides, the mass location information on social media shared by many users contains rich information on travel behavior of tourists and travel routes. Through integrating the group knowledge from the users, mining geotagged social media big data can discover popular tourist attractions and travel routes, and provide the more precise travel recommendation service. However, the existing research on big data-based travel recommendation has some problems such as lack of mining on multi-source information, ignoring the tourist context and so forth. Therefore, this project intends to make comprehensive use of multi-source social media, to dig out the tourists’ preference from their spatio-temporal trajectories, and especially taking the context of tourists into account, to set up a set of methods for context-aware personalized travel recommendation based on geotagged social media big data. Finally, the project attempts to take geotagged photos on Flickr and check-in data on Jiepang as data sources, and to build a travel recommendation system for method validation. This study can help tourists arrange the travel plan, improve their travel experience and quality. It also benefits destination development and marketing. Moreover, the study would stimulate the innovation of tourism industry, and promote its evolution toward smart tourism.

互联网时代下，旅游推荐是帮助游客过滤纷繁复杂的信息、做出合理旅行决策的重要手段，已日渐成为旅游领域的研究热点。同时，社交媒体上由众多用户所分享的海量的地理位置信息，蕴含着丰富的游客行为和旅游线路特征。通过对社交媒体地理大数据的挖掘，可以整合用户的群体知识，发现热门景点及旅行线路，提供更加智能的精准旅游推荐服务。然而，现有基于大数据的旅游推荐研究还存在对多源信息的挖掘不够、忽略游客所处情境等问题。因此，本课题拟综合利用多源社交媒体地理数据，根据游客的时空轨迹挖掘游客偏好，并充分考虑游客所处情境，建立一套基于社交媒体地理大数据的可感知情境的个性化旅游推荐方法，并以Flickr地理标签照片库与街旁网签到数据为基础，建立旅游推荐系统开展方法验证。本研究可帮助游客更好地安排旅行计划，提升旅游体验和品质；有利于旅游目的地开发与营销，同时也对激励旅游产业创新、传统旅游迈向“智慧旅游”有着积极的推动意义。

人们在旅行时往往习惯于在社交媒体上拍照和上传，由此形成了海量的含地理标签的照片数据（Geo-tagged photos），这些社交媒体照片数据包含了丰富的用户历史行为和用户信息。基于这些含地理标签的社交媒体数据，结合多源数据集（如天气数据、交通数据、文本数据等），可以很好地描述用户浏览历史情境，从而精细地刻画用户的“兴趣肖像”。然而，现有基于大数据的旅游推荐研究还存在对多源信息的挖掘不够、忽略考虑游客所处情境等问题。因此，本项目综合利用多源社交媒体地理数据，根据游客的时空轨迹挖掘游客偏好，并充分考虑游客所处情境，建立一套基于社交媒体地理大数据的可感知情境的个性化旅游推荐方法，并以Flickr地理标签照片库为数据基础，建立旅游推荐系统开展方法验证。本研究可帮助游客更好地安排旅行计划，提升旅游体验和品质；有利于旅游目的地开发与营销，同时也对激励旅游产业创新、传统旅游迈向“智慧旅游”有着积极的推动意义。