基于群体投票的移动性数据驱动地点类别推测  

Mobility Data-driven Location Type Inference Based on Crowd Voting

作　　者：熊可钦 阮思捷 杨芊雨 徐常炜 袁汉宁 XIONG Keqin;RUAN Sijie;YANG Qianyu;XU Changwei;YUAN Hanning(School of Computer Science and Technology,Beijing Institute of Technology,Beijing 100081,China;KQ GEO Technologies,Beijing 100023,China)

机构地区：[1]北京理工大学计算机学院,北京100081 [2]苍穹数码技术股份有限公司,北京100023

出　　处：《计算机科学》2025年第3期169-179,共11页Computer Science

基　　金：国家自然科学基金(62306033,42371480);教育部产学合作协同育人项目(231001223194844)。

摘　　要：地理信息是经济社会发展所需的基础数据,而兴趣点数据是其中一种常见且重要的数据类型。兴趣点数据的采集,传统上由地图厂商完成,存在成本高、空间覆盖不全、粒度不够细等问题,影响了下游应用的精准性。幸运的是,移动互联网的普及产生了大量移动性数据,其揭示了兴趣点的存在且具有推测地点类别的潜力。但是利用移动性数据推测地点类别因用户访问地点稀疏、移动上下文依赖关系复杂、用户的个体行为随机等问题面临挑战,现有工作无法较好地应对。因此,提出了一种基于群体投票的移动性数据驱动地点类别推测方法Milotic。该方法对地点类别的推测细化到每一条移动轨迹中,通过图模型建模了地点间复杂关系,通过签到嵌入和Bi-LSTM充分保留并融合了细粒度轨迹上下文信息,同时通过投票机制克服了个体行为的随机性。实验结果表明Milotic在两个真实移动性数据集上的加权F1值分别比最优基线提高了7.5%和13.3%。Geographic information serves as fundamental data for economic and social development.One of the common and vital type of data in this field is point-of-interest(POI)data.Previously,POI data are collected by map manufacturers,which are costly,have limited spatial coverage,and are not fine-grained enough,affecting the effectiveness of downstream applications.Fortunately,the popularization of the mobile Internet has generated vast amounts of mobility data that reveal the existence of POIs and have the potential to infer their location types.However,such potentiality is challenged by sparse visited locations by users,complex contextual dependency,and random individual behaviors,which are not adequately addressed by existing work.Therefore,we propose a mobility data-driven location type inference method based on crowd voting,namely Milotic.This method refines the task of predicting location types to each trajectory,models complex relationships between locations with graph models,fully retains and integrates fine-grained trajectory context information through check-in embeddings and Bi-LSTM,and overcomes the randomness of individual behaviors through a voting mechanism.Experimental results demonstrate that Milotic achieves weighted F1 score improvements of 7.5% and 13.3% respectively over the best baseline on two real-world mobility datasets.

关 键 词：时空数据挖掘 自发地理信息 地点类别推测 移动性数据 兴趣点 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]