CRUISEResearchGroup/TripWorld

Trip World是一个公开可重分发的、跨城市旅行基准数据集，包含大量异地旅行序列、丰富的Foursquare和Google地图兴趣点（POI）元数据，以及统一的Google评论语料库。该数据集旨在支持轨迹推荐、跨城市POI排序和下一个POI预测、地点建模（基于1.86亿条Google评论和68.7万个POI的结构化属性），以及旅游和移动性分析（覆盖90多个国家1,173个城市的旅行流分析）。数据集基于STD-2018签到流构建，经过城市合并、家乡发现、旅行者过滤和POI增强处理。核心统计数据包括：518,567条旅行行为记录、148,402名独立旅行者、687,173个独立POI、1.86亿条评论，时间跨度为2017年10月3日至2018年10月20日。数据集采用Apache Parquet格式，包含旅行行为、POI摘要、区域标签、元数据、地点属性和评论等多个文件，并应用了用户ID重映射、评论者身份匿名化等隐私保护措施。Trip World is a publicly redistributable, cross-city travel benchmark that pairs a large corpus of out-of-town travel sequences with rich Foursquare and Google Maps point-of-interest (POI) metadata and a unified Google reviews corpus. The dataset is designed to support trajectory recommendation, cross-city POI ranking and next-POI prediction, place modelling (driven by 186 million Google reviews and structured attributes for 687K POIs), and tourism and mobility analytics (aggregate analyses of travel flows between 1,173 cities in 90+ countries). It is built on top of the STD-2018 check-in stream, with metro-consolidation, hometown discovery, traveller filtering, and POI enrichment applied. Key statistics include: 518,567 travel-behavior records, 148,402 distinct travellers, 687,173 distinct POIs, 186 million reviews, and a time span from October 3, 2017 to October 20, 2018. The dataset is provided in Apache Parquet format, with files for travel behaviors, POI summaries, region labels, metadata, place attributes, and reviews, and includes privacy measures such as user ID remapping and reviewer identity anonymization.