Research and application of intervening opportunity class models for predicting human mobility

The research on human mobility has been an important topic in many fields such as transportation science, sociology and economic geography. With the fast development of urbanization and globalization, human travel has become increasingly complex. Understanding and predicting human mobility is of great theoretical significance to explain various complex phenomena related to human activities and has practical applications in many fields. For example, in terms of urban planning, the research on human mobility can help us better understand the travel demand and travel characteristics of urban residents, so as to optimize the urban road network and improve the efficiency of the public transport system; in terms of migration, the research on human mobility can help us better understand migration patterns, so as to provide a scientific basis for the government to develop population management policies; in terms of the spread of diseases, the research on human mobility can help us better understand the travel patterns and travel characteristics of different groups of people, so as to more accurately predict and control the spread of disease within humans. Through the mining of large-scale human travel data, this thesis takes the trip distribution prediction as the main research object, uses the sys- tem science thinking and combines data statistical analysis methods to establish a model that can predict human mobility at different spatiotemporal scales, and further explores its application on spatial interaction and enriches the equipotential line display method of model prediction results. The main results of this thesis are as follows:
(1) We collect a variety of mobility datasets at different spatiotemporal scales, including commuting datasets, migration datasets, freight datasets, intracity trip datasets and intercity travel datasets, providing a reliable data basis for the research of human travel behavior. Based on these datasets, we count travel distance distribution, travel fluxes distribution and normalized entropy, and show the spatial distribution characteristics by drawing the map of trip generation, trip attraction and desire line.
(2) We propose a new interventional opportunity class model named universal opportunity model to predict commuting, migration, freight, intracity trip and intercity travel, which has broad applicability. The universal opportunity model assumes that individuals will comprehensively consider the benefits of each location when choosing a destination, and it reflects two human behavioral tendencies: one is the exploratory tendency and the other is the cautious tendency. The exploratory tendency describes the individual tend to choose the destination whose benefit is higher than the benefits of the origin and the intervening opportunities, and the cautious tendency describes the individual tend to choose the destination whose benefit is higher than the benefit of the origin, and the benefit of the origin is higher than the benefits of the intervening opportunities. Validation results on multiple datasets show that the universal opportunity model can better predict human mobility than previous intervening opportunity class models. The universal opportunity model establishes a new framework in intervening opportunity class models and covers the classical radiation model and opportunity priority selection model, which can help us better understand the underlying mechanism of the individual’s destination selection behavior.
(3) By extending the universal opportunity model, we propose an interactive city choice model to measure the interaction intensity between cities and city interaction intensity. The model assumes that the probability of an individual choosing to interact with a city is proportional to the number of opportunities in the destination city and inversely proportional to the number of intervening opportunities between the origin city and the destination city, calculated using the travel time in the transportation network. By using this model to measure the interaction intensity and combining the outgoing interaction in- tensity and the incoming interaction intensity to quantify city interaction intensity, we an- alyze the impact of changes in the Chinese land transportation network from 2005 to 2018 on the intercity and city interaction intensity. The results show that the travel time between cities has decreased and the interaction intensity between large cities has increased due to the development of land transportation. In particular, the interaction intensity of cities along high-speed railways has greatly increased. Compared with the gravity model and the radiation model, the interactive city choice model can better measure the interaction intensity between cities and city interaction intensity.
(4) We propose a new method to display the prediction results of the universal opportunity model by establishing a vector field and drawing the equipotential lines. Compared with the previous commuting vector field method, our method is more general and can be applied not only to commuting but also to passenger travel and freight travel. Further, we expand it and propose a generalized vector framework, which includes two indicators: the vector field and track orientation. Among them, the vector field can be used to describe the characteristics of the field formed by individuals’ trajectories, and the track orientation divides the trajectory into positive, negative and balanced according to whether the individual tend to move toward the starting point or away from it during the trip process. Through the analysis of freight datasets in 21 Chinese cities and Foursquare check-in datasets in New York, we find that there are more trips point to the starting point than away from the starting point, and the ratio of trips away from the starting point and trips back to it is similar, which indicates that there is a universal rule in different cities. Finally, we introduce the distance rand model, the distance traveling salesman model and the distance rand-traveling salesman mix model to generate the trajectory, and apply the generalized vector framework to study the characteristics of the trajectory generated by these models. The results show that the distance rand-traveling salesman mix model can reproduce the universal rule that the ratio of trips away from the starting point and trips back to it is similar in different cities. The above research shows that the generalized vector field framework can not only display the spatial characteristic of population mobility, but also quantify the spatial characteristic of individual mobility.