Modeling And Spatio-Temporal Detection Of Short-Term Out Of Service Behavior Based On Taxis' GPS Trace Big Data

China's urbanization is in a stage of rapid development. Large amount of "urban diseases" that urban resources allocation cannot meet the demands of urban activities are emerging. Taxis play an important role in urban public transportation resources. There are over 60 thousands taxis in Beijing City, about 20 thousands in Wuhan City, and about 16 thousands in Shenzhen City, currently. Taxis make up quite proportion in urban vehicles. Taxi drivers' out-of-service stay behaviors such as having meals, refueling and taking a rest reflect the spatiotemporal operating efficiencies of taxi driver group as well as taxi drivers' demands on the out-of-service infrastructure allocation. Taxis' spatiotemporal GPS (Global Positioning System) trajectories are one kind of positioning big data with high coverage and low collecting cost. Existing researches based on taxis' GPS trace data have not taken the characteristics and demands of taxi drivers' out-of-service behaviors into consideration. As a result, based on taxis' spatiotemporal GPS trace data, studying the spatiotemporal distribution characteristics of taxi drivers' refueling behaviors, detecting spatiotemporal distribution rules of taxi drivers' refueling behaviors and uncovering the adaptation between urban group activities and resources allocation are of important theoretical and application value.This paper realizes recognition and detection of taxis' refueling activities through analyzing the spatiotemporal characteristics of taxis' refueling activities, modeling taxis' refueling activities and analyzing their spatiotemporal distribution using taxis'GPS trace big data. This paper puts forward a point-line feature cross K- function method for studying the spatial correlation between refueling activities and refueling stations POIs. By the proposed approaches, adaptation and matching degree between urban group activities and resources allocation are discovered and scientific evidences for urban resources allocation optimization are provided. The main studying content are listed below:(1) By analyzing spatial and temporal characteristics of taxis' refueling activities, this paper presents a modeling approach for taxis'refueling activities and realizes effective detection of taxis'refueling activities in Wuhan City, with the trace data collected from 10614 taxis in Wuhan.(2) This paper puts forward a linear event description for taxis' refueling activities based on analyzing of the linear features of taxis' refueling activities. This paper analyzes and describes the spatiotemporal distribution characteristics of taxis' refueling group activities in Wuhan City accurately with Kernel Density Estimation for linear features in planar space.(3) This paper proposes a point-line cross K-function correlation analyzing method, realizing the measurement of the correlation between the spatial distribution of urban refueling stations POIs and taxis' refueling linear events in different distance scales. The proposed method is able to discover the matching degree between taxis' refueling group activities and the allocation of urban refueling stations public resources.