Date Published: February 12, 2018
Publisher: Public Library of Science
Author(s): Jungmin Kim, Juyong Park, Wonjae Lee, Feng Xia.
The quality of life for people in urban regions can be improved by predicting urban human mobility and adjusting urban planning accordingly. In this study, we compared several possible variables to verify whether a gravity model (a human mobility prediction model borrowed from Newtonian mechanics) worked as well in inner-city regions as it did in intra-city regions. We reviewed the resident population, the number of employees, and the number of SNS posts as variables for generating mass values for an urban traffic gravity model. We also compared the straight-line distance, travel distance, and the impact of time as possible distance values. We defined the functions of urban regions on the basis of public records and SNS data to reflect the diverse social factors in urban regions. In this process, we conducted a dimension reduction method for the public record data and used a machine learning-based clustering algorithm for the SNS data. In doing so, we found that functional distance could be defined as the Euclidean distance between social function vectors in urban regions. Finally, we examined whether the functional distance was a variable that had a significant impact on urban human mobility.
The latest models used to predict intra-city traffic employ physical factors, such as the population size and the distance-of-travel, as driving variables. However, because of the increasing complexity of cities, new variables have been suggested that would improve the predictive power of these models. The social function of inner regions was one such variable.
Existing studies have defined the function of regions by using land use [51, 52] or social demographics [53–55]. However, the land use classification method is often limited by the increase in urban complexity, whereas social demographic studies have only focused on the details of individual properties. For our purpose, it was necessary to systematically classify the function of urban regions while showing the functional complexity of a specific area. Fortunately, due to quantitative and qualitative increases in urban data, the possibility of precisely defining the functions of urban regions has risen. Based on this, we aimed to present integrated functions that reflected more detailed characteristics of urban regions. In this section, we have expounded on defining the function of urban regions with various variables.
As mentioned before, the function of urban regions was difficult to express with some variables. To verify the hypothesis that the functions of urban regions affected human mobility in the city, it was necessary to include numerous values that represented the functions of urban regions in our human mobility prediction model. However, the number of public records in this study was 790, and when variables from SNS data analysis were included, it was 810. Also, if this data would have been handled like the Mass term of a gravity model, the number of variables to considered would have doubled. Moreover, considering the continuous increase in the number of data types that could be used to reflect characteristics of the regions, simply adding all variables into the model did not seem appropriate. Of course, it was not entirely impossible to predict human mobility through models with a large number of variables, but we wanted to find a better solution.
Table 6 (1)-(2) show that significant positive correlations existed between the two different functional distances and the traffic volume determined as a result of regression analysis of the modified gravity model. In other words, distance and mass conditions were controlled, which meant that the farther the functional distance was, the greater the traffic volume. This indicated that the functional distance between regions could be used as another variable for predicting urban traffic volume.
In this study, we verified the best-use terms for a gravity model to determine urban human mobility. With the exception of the floating population (which has a tautological problem), the number of tweets was found to be the best value for predicting urban human mobility for the mass term. However, all mass values were used at the same time because all mass values were considered to be significant even if their explanatory power values were different.