Abstract
People in cities nowadays suffer from increasingly severe traffic jams due to less awareness of how collective human mobility is affected by urban planning. Besides, understanding how region functions shape human mobility is critical for business planning but remains unsolved so far. This study aims to discover the association between region functions and resulting human mobility. We establish a linear regression model to predict the traffic flows of Beijing based on the input referred to as bag of POIs. By solving the predictor in the sense of sparse representation, we find that the average prediction precision is over 74% and each type of POI contributes differently in the predictor, which accounts for what factors and how such region functions attract people visiting. Based on these findings, predictive human mobility could be taken into account when planning new regions and region functions.
Highlights
People in cities suffer from increasingly severe traffic jams
This study aims to answer two questions that have never been explored: (1) Do certain region functions lead to predictable human mobility, and is it possible to predict such human mobility from region functions? (2) Is there any deterministic relationship between each type of region function and the attraction of people? If so, how does one quantitatively evaluate the impact factor of every causal of human mobility? In this study, we format the problem as sparse representation rendered prediction and optimize the predictor in a manner similar to variable selection
The physical meanings of the dominant weights are explained though the 3 examples as follows: It is known that people usually leave home for office in the morning rush hour, travel between different work places during the day and return home or go out for entertainment in the evening
Summary
People in cities suffer from increasingly severe traffic jams. Tremendous efforts have been made to study the mechanism regarding how traffic jams are generated. The endeavor being devoted to human mobility pattern mining and traffic flow prediction has resulted in the following analytical methodologies: Statistical Methods [2][3], Nonnegative Matrix Factorization and Optimization Methods [4], Entropy-maximizing Methods [5], Multiscale Radial Basis Function (MSRBF) networks [6], and Deep Learning Methods [7]. Such studies are limited in that they do not take into account the social contexts in which traffic flows are generated, say, the motivations that drive people to travel from one place to another place.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
