Recovering Individual’s Commute Routes Based on Mobile Phone Data

Abstract

Mining individuals’ commute routes has been a hot spot in recent researches. Besides the significant impact on human mobility analysis, it is quite important in lots of fields, such as traffic flow analysis, urban planning, and path recommendation. Common ways to obtain these pieces of information are mostly based on the questionnaires, which have many disadvantages such as high manpower cost, low accuracy, and low sampling rate. To overcome these problems, we propose a commute routes recovering model to recover individuals’ commute routes based on passively generated mobile phone data. The challenges of the model lie in the low sampling rate of signal records and low precision of location information from mobile phone data. To address these challenges, our model applies two main modules. The first is data preprocessing module, which extracts commute trajectories from raw dataset and formats the road network into a better modality. The second module combines two kinds of information together and generates the commute route with the highest possibility. To evaluate the effectiveness of our method, we evaluate the results in two ways, which are path score evaluation and evaluation based on visualization. Experimental results have shown better performance of our method than the compared method.