公共自行车站点租借规律预测研究与应用

发布时间：2018-10-20 06:50

【摘要】：随着城市化的发展,出现了一些交通和环境问题,它们成为困扰很多大城市发展的重要问题。公共自行车系统(Public Bicycle System)也叫PBS,它的出现有效地改善了交通问题。但是在PBS的使用过程中,由于居民出行的“潮汐性”所产生的“没有剩余车辆可借、没有空闲车桩可还”问题,这个问题被简称为“借车难、还车难”问题,它对PBS的发展存在着严重的制约和影响。因此为了解决这个问题,研究并利用相关算法及时准确预测出可租借的公共自行车数量,并及时地通知给用户最佳租还车站点成了当务之急。目前针对公共自行车交通流预测模型的研究较少,但是针对机动车在相关方面的研究却已经十分成熟,因此,目前针对公共自行车交通流预测的研究主要还是借鉴了机动车相关方面的研究。由于公共自行车交通流与机动车交通流的差异性,目前的相关研究并不能精确的预测公共自行车交通流,尤其是在可租借公共自行车数目长短期预测研究上。本文通过对公共自行车使用特性和租借规律的分析研究,提出了一种结合“天气、气温、节假日”三种因素的未来可租借公共自行车长时租借预测模型以及利用粒子群优化支持向量机模型(PSO-SVM)的短时租借预测模型。本文通过长时租借预测模型,为调度决策者进行提前调度提供理论依据。通过短时租借预测模型,可以准确的预测短期内可租借公共自行车数目。本文的主要研究思路如下:(1)分别从租借行为特性分析,出行时间特性分析,空间行为特性分析三个方面分析归纳各种类型公共自行车站点的租借规律,归纳出各类公共自行车站点的租借规律,总结出哪类站点在什么时间容易出现“借车难、还车难”的问题。(2)从PBS中获得用户租借记录,并且结合互联网上公布的天气数据,气温等数据,利用k均值算法和NB分类预测算法,结合曲线拟合技术建立了公共自行车租借规律预测模型,对未来站点内可租借公共自行车可租借数目进行长期预测分析,实验结果可以作为调度决策者调度车辆的理论依据。(3)本文提出了对于公共自行车可租借数目利用PSO-SVM模型对公共自行车可租借数目进行短期预测,实验结果说明该模型具有良好的精度。
[Abstract]:With the development of urbanization, some traffic and environmental problems have emerged, which have become an important problem that puzzles the development of many big cities. The emergence of the public bicycle system (Public Bicycle System), also known as PBS, has effectively improved traffic problems. However, during the use of PBS, the problem of "no surplus vehicles to borrow, no free car piles to return", which is caused by the "tidal nature" of residents' travel, has been referred to as the problem of "difficult to borrow a car, to return a car". It has serious restriction and influence on the development of PBS. In order to solve this problem, it is urgent to study and use relevant algorithms to predict the number of public bicycles that can be rented in time, and to notify users of the best rental and return stations in time. At present, there are few researches on the prediction model of public bicycle traffic flow, but the research on the related aspects of motor vehicles has been very mature, so, At present, the research on the prediction of public bicycle traffic flow is mainly based on the research of motor vehicles. Due to the difference between the traffic flow of public bicycle and the traffic of motor vehicle, the current related research can not predict the traffic flow of public bicycle accurately, especially in the long-term and short-term prediction of the number of public bicycle. Based on the analysis of the characteristics of public bicycle use and the law of leasing, this paper proposes a combination of "weather, air temperature," The prediction model of future long-term rental of public bicycle with three factors and the short-term rental prediction model using particle swarm optimization support vector machine (PSO-SVM) model are discussed in this paper. This paper provides a theoretical basis for scheduling decision makers to advance scheduling through long term lease prediction model. The short-term rental prediction model can accurately predict the number of public bicycles that can be leased in the short-term. The main research ideas of this paper are as follows: (1) the rental law of various types of public bicycle stations is analyzed and summarized from three aspects: (1) analysis of rental behavior characteristics, analysis of travel time characteristics, and analysis of spatial behavior characteristics. This paper sums up the leasing rules of all kinds of public bicycle stations, and summarizes which stations are prone to the problem of "difficult to borrow a car, difficult to return a car" at what time. (2) obtaining user rental records from PBS and combining with the weather data published on the Internet, Based on the data of air temperature and NB classification prediction algorithm and curve fitting technique, the prediction model of public bicycle rental law is established, and the long-term prediction analysis of the number of public bicycle leasehold in future stations is carried out. The experimental results can be used as the theoretical basis for scheduling decision makers. (3) this paper proposes a short-term prediction of the number of public bicycles available for leasing by using PSO-SVM model. The experimental results show that the model has good accuracy.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;U491

【参考文献】