Urban Hotspots Detection And Residents’ Travel Behavior Analysis In Small And Meduim-Sized Cities Based On Shared E-Bike Trajectory Data Mining

Traditional urban spatial planning focuses on static land expansion and ignores the influence of urban residents’ behavior on urban spatial structure,thus leading to a series of urban diseases such as traffic congestion and resource waste.Constructing peopleoriented city is the goal of urbanization development in the future.Trajectory data contains rich human-region interaction information,which provides an opportunity to explore the impact of residents’ behavior on urban sturcture.Shared electric bikes,an emgering transportation,have the characteristics of low cost,fast speed,free travel and high accessibility,which make them an excellent solution for short-and medium-distance trips in small and medium-sized cities.Due to the limits of the bicycle battery and positioning device signal,the raw data quality of shared e-bikes is inherently sparse,characterized by noncontiguous,heterogeneous GNSS points and a lack of riding status information,which bring challenges to the mining of residents’ behavior information.In this content,based on the trajectory of shared electric bikes for the first time,this dissertation studies the technology of trajectory data cleaning and urban hotspot extraction and excavates residents’ travel behavior in small and medium-sized cities,which provides a strong technical support for promoting the harmonious development of small and medium-sized cities.The main research contents and conclusions are as follows.(1)The trajectory of shared electric bicycles is is inherently sparse,characterized by noncontiguous,nonuniform GNSS points and a lack of riding status information,which leads to difficulties or low precision in trajectory cleaning methods.Inspired by the velocity sequence linear clustering(VSLC)algorithm,a trajectory cleaning method of multi-rule constrained homomorphic linear clustering(MCHLC)is proposed.This algorithm introduced direction constraint and context semantic constraint criterion to improve the performance,which not only effectively reduced the interference of pseudo stop in the identification of stay points caused by residents’ accidental behavior,but also effectively identified the temporary stop with specific purposes or behaviors,thereby revealing the details of residents’ travel behaviors.Compared with the baseline VSLC algorithm,the performance of MCHLC algorithm is improved by approximately 10%,reaching 95.62%.Compared with the baseline CB-SMo T(Clustering Based Stop and Move of Trajections)algorithm,the MCHLC algorithm is not affected by number of points in stopping point recognition,and both "single point" stops and "multiple points" stops can be identified.(2)A Gaussian mixture model based on density peak clustering(DPC-GMM)method is proposed to detect urban hotspots,aiming to solve the issue of difficult parameter setting,thereby adapting to the uneven spatial distribution of trajectory data.In this method,the number and the centers of the clusters are automatically determined by the decision graph of density peak,and the parameters automatically are suitalble to the data spatial distribution through Gaussian mixture model,which effectively solves the issue of parameter setting.Compared with traditional hot spot detection methods,DPC-GMM method is a multi-scale hot spot detection method,and has obvious advantages in "weak" hot spot detection.In addition,the DPC-GMM method can also determine the core and boundary of hot spots.The hot spots detection of DPCGMM method have a high degree of center aggregation.(3)Taking Tengzhou City as an example,the travel behavior of residents in small and medium-sized cities is investigated from multiple perspective based on the trajectory data of shared electric bicycles.The CARA(Commuting Activity and Residential Areas)and CAWZ(Commuting Activity and Working Zones)models are established to quantitatively describe the interaction between "residents’ behavior and activities" and "urban functional areas".The urban commuting space is extracted and the commuting efficiency of residents is evaluated by the excess commuting theory.The results are as follows:(1)The travel distance of shared electric bikes is concentrated within 5km,the travel time is between 5 and 10 min,and the cycling speed is between 10km/h and 20km/h.(2)Shared electric bicycle travel is closely related to weather conditions,temperature and air quality.(3)The residents’ travel behavior pattern in Tengzhou is characterized with a significant "multi-peak",which is different from the "double-peak" travel behavior of big cities.The travel pattern of weekday and weekend has obvious differences,showing significant temporal characteristics.(4)The CARA model is an increasing logarithmic function,indicating that urban hotspots have positive correlation with residential areas.The CAWZ model is a piecewise function,in which the relationship between the middleand low-hotspots and the work zones satisfies the Gaussian function,and the relationship between the high hotspots and the work zones follows the decreasing logarithmic function.(5)The spillover commuter index of Tengzhou is 11.3%,which proves that there is no obvious job-housing separation phenomenon,indicating that residential areas are closely connected with work zones and residents’ commuting efficiency is high.There are 72 figures,15 tables and 181 references in this dissertation.