Finding Spatiotemporal Patterns of Mobile Application Usage
Tong Li,Yong Li,Tong Xia,Pan Hui
DOI: https://doi.org/10.1109/tnse.2021.3131194
IF: 6.6
2024-01-01
IEEE Transactions on Network Science and Engineering
Abstract:Understanding mobile application usage patterns is significant for producing better services and enriching user experience. The understanding of spatiotemporal patterns of application usage is still limited. In this paper, we aim at finding spatiotemporal mobile app usage patterns and propose a framework to capture who, when, where, and what applications are used. We first collect a large-scale and real-world application usage dataset covering over 400 thousand active users and 600 million records. In order to introduce spatial features, we partition the collection area into small regions. By grouping regions of similar point-of-interest attributes, we then map 796 regions onto 13 region clusters with semantic meanings. As a result, the original data is reformed as a tensor of four dimensions, i.e., users, application categories, region clusters, and time-slots. We then leverage a multi-way clustering algorithm on the tensor to extract coupling relations between different dimensions. Finally, we discover 508 distinct spatiotemporal application usage patterns with meaningful labels, including E- readers, Digital payers, Uber/DiDi drivers, Young parents, Travellers and Travel planners. The results produced by our framework can serve a series of applications, e.g., identifying user habits and inferring user demographics, which are helpful in customized services like personalized recommendations.