基于出租车GPS数据的用户出行热点挖掘与交通流波动分析

发布时间：2018-05-03 15:05

本文选题：OD预测 + 非负矩阵分解　；参考：《西南大学》2017年硕士论文

【摘要】：随着中国城镇化进程日益加快,城市规模急剧扩张,居民出行方式不断变化,居民出行范围逐步扩大,但由于城市地理位置的限制及交通基础设施的建设滞后,从而导致一系列制约城市经济发展和居民生活水平提高的城市交通问题,包括交通拥堵、资源分配不均等。研究城市居民出行模式和交通流量波动现象为解决出租车空载率过高、居民出行需求无法满足、交通管理效率较低等交通问题提供可能性。鉴于此,本文通过对出租车GPS数据的分析,挖掘居民出行规律及区域交通流的波动现象,并提出基于非负矩阵分解的自回归预测模型和实现对交通流波动现象的定量分析,为交通使用者提供实时有效的出行信息和有效的交通流量波动演化规律,从而对缓解目前的交通问题提供实质性地帮助及建议。面对日益严重的城市交通问题,本文通过挖掘和分析海量出租车GPS轨迹数据,基于矩阵分解算法和时间序列模型实现对起讫(Origin-Destination,OD)矩阵的预测。并通过对流量粗粒化建模及复杂系统波动理论分析,本文已实现对区域网络单节点流量的动态分析、全节点流量的波动规律以及区域网络系统内外部流量的分离分析。本文主要贡献如下:(1)提出基于非负矩阵分解的自回归模型,主要通过对OD矩阵的非负特征和可模拟的用户出行特征,本文引入非负矩阵分解算法对用户出行特征进行分析和宏观描述;同时,在此基础上利用自回归模型对OD矩阵进行预测和估计。(2)基于北京市出租车交通数据,实现基于非负矩阵分解的自回归(Nonnegative Matrix Factorization-Auto Regressive,NMF-AR)模型对OD矩阵的预测。基于出租车GPS数据,通过NMF-AR模型挖掘和预测用户出行信息,同时与引入的短时交通流预测模型进行对比分析,并且对预测精度、模型参数、数据敏感度等问题进行深入分析,验证模型的预测能力。通过对OD矩阵实时预测分析可为居民提供实时有效的出行信息,可有效降低空载率,提高运营效益。(3)实现基于流量的粗粒化建模和深入分析交通流波动情况。在北京市区域棋盘式划分策略和区域交通网络的基础上,本研究对北京市出租车GPS数据进行适当的数据预处理工作。一方面,针对单个区域的交通流量波动情况,本文利用粗粒化方法处理流量变化构建对应的网络,同时分析节点车流量的波动情况。另一方面,本文对区域网络交通流量波动的规律和演化进行重点分析,并对多个区域交通流量的总体特性进行考察,并实证分析该波动规律。基于北京市出租车GPS轨迹数据,本文从真实区域出租车流量实证分析其基于时间的流量均值和标准差之间的关系。同时通过对区域网络内外部流量的分解分析,发现网络系统流量波动的演化规律。从而为交通管理部门优化和制定交通管理策略有效信息。综上,通过用户出行OD矩阵建立NMF-AR预测模型、基于区域流量的粗粒化分析及交通流幂律现象定量分析,可为交通运营提供实时的出行规律,为交通监管部门提供有效的建议,从而有助于降低出租车空载率、优化交通应急管理、提升城市交通运行效率。
[Abstract]:With the rapid urbanization process in China, the urban scale is expanding rapidly, the mode of resident travel is changing, and the residents' travel scope is gradually expanding. However, because of the limitation of urban geographical location and the lagging of the construction of traffic infrastructure, a series of urban traffic problems which restrict the development of urban economy and the improvement of the living standard of residents are caused. Including traffic congestion and uneven distribution of resources. The study of urban residents' travel mode and traffic flow fluctuation is a possibility to solve the problem of overloading of taxi no-load, the inability to meet the needs of the residents and the low efficiency of traffic management. In view of this, this paper makes an analysis of the GPS data of the taxi and excavates the laws and areas of the residents' travel. The fluctuation phenomenon of the traffic flow is presented, and the autoregressive prediction model based on the non negative matrix decomposition and the quantitative analysis of the fluctuation of traffic flow are put forward to provide the traffic users with real time effective travel information and the effective evolution law of the traffic flow fluctuation, thus providing substantial help and suggestion to alleviate the current traffic problems. Facing the increasingly serious urban traffic problems, this paper realizes the prediction of the Origin-Destination (OD) matrix based on the matrix decomposition algorithm and the time series model by mining and analyzing the mass taxi GPS trajectory data. Through the rough modeling of the traffic and the analysis of the wave theory of the complex system, this paper has realized the regional network single. The dynamic analysis of node flow, the fluctuation law of the full node flow and the separation and analysis of the internal and external flow of the regional network system. The main contributions of this paper are as follows: (1) the autoregressive model based on the non negative matrix decomposition is proposed. The non negative matrix decomposition algorithm is introduced in this paper mainly through the non negative characteristics of the OD matrix and the simulated user travel characteristics. The characteristics of user travel are analyzed and macroscopically described; at the same time, the OD matrix is predicted and estimated by the autoregressive model. (2) based on the Beijing taxi traffic data, the prediction of the OD matrix based on the Nonnegative Matrix Factorization-Auto Regressive (NMF-AR) model based on the non negative matrix decomposition is based on the prediction of the OD matrix. Taxi GPS data, through NMF-AR model mining and prediction of user travel information, and compared with the introduction of short-term traffic flow prediction model, and in-depth analysis of the prediction accuracy, model parameters, data sensitivity and other issues to verify the prediction ability of the model. Through the real-time prediction analysis of the OD matrix can provide the residents with real information. The effective travel information can effectively reduce the no-load rate and improve the operation efficiency. (3) realize the rough modeling based on the traffic flow and analyze the fluctuation of traffic flow. On the basis of the regional chessboard division strategy and regional traffic network in Beijing, this study carries out the appropriate data preprocessing to the taxi GPS data in Beijing. In view of the fluctuation of traffic flow in a single region, this paper uses the coarse graining method to deal with the flow change and constructs the corresponding network, and analyzes the fluctuation of the node traffic flow. On the other hand, this paper focuses on the law and evolution of the regional network traffic flow fluctuation, and the overall characteristics of the traffic flow in multiple regions. Based on the taxi GPS trajectory data of Beijing City, this paper empirically analyses the relationship between the mean and standard deviation of the time based traffic flow from the real area taxi traffic flow. At the same time, through the analysis of the internal and external flow of the regional network, the evolution of the flow fluctuation of the network system is found. On the other hand, the traffic management department optimizes and establishes the effective information of traffic management strategy. To sum up, the NMF-AR prediction model is established by the OD matrix of user travel. Based on the roughing analysis of regional flow and the quantitative analysis of the power law phenomenon of traffic flow, it can provide real time travel rules for traffic operation and provide effective suggestions for traffic supervision departments. It helps to reduce the empty load rate of taxis, optimize traffic emergency management and improve the efficiency of urban traffic operation.

【学位授予单位】：西南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491

【参考文献】