Leveraging Individual and Collective Regularity to Profile and Segment User Locations from Mobile Phone Data

Abstract

The dynamic monitoring of home and workplace distribution is a fundamental building block for improving location-based service systems in fast-developing cities worldwide. Inferring these places is challenging; existing approaches rely on labor-intensive and untimely survey data or ad hoc heuristic assignment rules based on the frequency of appearance at given locations. Motivated by the regularities in human behavior, we propose a novel method to infer the home, workplace, and third place based on an individual’s spatial-temporal patterns inferred from Call Detail Records. To capture the individual regularity, our method develops, for each person-location, the probability distribution that the person will appear in that location at a specific time of day using geo-temporal travel patterns a panel of individuals. To reveal the collective regularity, we apply eigen-decomposition to the matrix of historical geo-temporal data. Unsupervised machine learning techniques are then used to extract commonalities across locations for different groups of travelers, making inferences, such as home and workplace. Testing the methodology on real-world data with known location labels shows that our method identifies home and workplace with significant accuracy, improving upon the best practices in the literature by 79% and 34%, respectively. The methodology proposed is computationally efficient and is highly scalable to other real-world applications with historical tracking data. It provides a basis to improve location-based services, such as mobile commerce, social events recommendations, and urban transit design.