Time Series Analysis And Complex Network Modeling Based On Public Bicycle Riding Data

In recent years,with the concept of green,healthy and shared travel were proposed,the public bicycle system has experienced a vigorous development.By the end of 2018,more than 1,000 cities around the world had established public bicycle systems.The system has the characteristics of green,low carbon and environmental protection.The system can effectively solve the problems of people's short-distance travel and transportation connection,relieve the high load pressure of urban transportation vehicles,reduce traffic congestion,automobile exhaust pollution and other problems.Although system has many advantages,some problems still bother consumers and managers.For consumers,due to the diversity of residents' travel demands,the distribution of bicycles in time and space is unbalanced,which leads to the problem of difficulty for users to use bicycles.For managers,because of the existing bicycle scheduling method has on time lag,so when the car arrived at the designated place,station requirements often have changed.The present study focuses on public bike systems and investigates the internal structure and network characteristics of public bike systems with the use of the theories and methods of statistics and network science,the details are as followsFirstly,this paper obtained the original data from Citi Bike's official website and preprocessed the data with python software.High quality data sets are obtained by removing incomplete,incorrect,and redundant data sets from the original data set.Through the statistical analysis of the processed data,we can understand the characteristics of the New York public bicycle system.Through programming,the time series of bicycle inflow and outflow in 24 moments of each station were obtained,so as to prepare the data for the subsequent stations clustering study.Secondly,a clustering study is carried out on the stations of the public bicycle system in order to explore the influence between the usage patterns and geographical locations of different types of stations.In this paper,the net outflow of normalized bicycle is proposed as a new clustering index.The clustering index adopted by the original research relies on the static data of stations and are not sensitive to stations faults,resulting in the deviation of the clustering result.The new clustering index can describe the use status of stations dynamically,and eliminate the errors caused by faulty vehicles and piles.K-means clustering method is used to cluster the time series data of the station,and the clustering results are displayed spatially.Thirdly,in the existing complex network modeling research on public bike system,some of them do not consider the direction or weight of the edge,and some ignore the time attribute of the edge.In this paper,the traffic weighted directed network is constructed by taking the public bicycle station as the network node,the cycling lines between the stations as the directed edges,and the passenger flow and person-time between the stations as the weight of the adjacent edges.Taking the public bike station as the network node,the cycling routes between the stations as the directed edges,and the average cycling time on each line as the weight of the adjacent edges,the time weighted directed network is constructed.Both of the two network models consider the direction of the edge,and consider the impact of traffic and time on the network model,which can reflect the actual situation of the network more truly.Since the network has the property of directed and weighted,the concept of strength difference distribution and the concept of unit strength distribution are proposed to describe the topological characteristics of the network.For different weighted directed networks,the strength difference distribution and the unit strength distribution can describe different practical meanings respectively.For traffic weighted directed network,the strength difference distribution can reflect the situation of traffic flow in and out of the station.The unit strength distribution describes the passenger transport capacity of a single edge.For the time weighted directed network,the unit strength distribution can analyze the time rule of residents' cycling.Finally,in the traffic weighted directed network,the connection between nodes will be disconnected and connected with the change of time,so the traffic weighted directed network belongs to a typical temporal network.By selecting different time window division methods,the timing characteristics of traffic weighted directional network are analyzed,and the changes of temporal network in one week and one day are understood.The relative size of the maximal connected subgraph and the average path length are selected as the robustness evaluation indexes to analyze the robustness of the traffic weighted directed network.