公共自行车租赁点车辆数的预测方法研究

发布时间：2018-04-15 20:07

本文选题：公共自行车系统 + 自行车数目预测　；参考：《南京师范大学》2015年硕士论文

【摘要】：公共自行车系统能有效的解决人们出行中“最后一公里”的问题,能更好地提升城市公共交通的整体服务水平。然而,“借车难,还车难”是公共自行车系统的主要问题之一,直接影响着用户的满意度。自行车调度是解决这些的有效方法之一,而租赁点的自行车数目预测是自行车调度的核心问题之一。因此,本文针对公共自行车系统特点,进行租赁点自行车数目预测研究,具有重要的理论意义与应用价值。(1)基于租赁点的自行车数目的变化规律分析,本文提出了一种新的租赁点自行车数目的预测框架。该框架包括结合数据选择,预测模型和误差补偿三部分,以预测模型为基础,结合误差补偿机制,能大大的提高预测的精度。(2)为了挖掘公共自行车各租赁点的变化规律,给预测模型提供良好的数据支持,本文提出了基于自行车使用规模的公共自行车租赁点的聚类方法。通过对不同天气(晴天、阴天、大雨、大雪、大雾、大风等)、季节、日期类型以及不同租赁点的分析,结合温州市鹿城区公共自行车系统的实际运行数据,对各种类型租赁点在不同外部条件下的变化曲线进行描述和分析,并将变化规律和幅度表示为特征串,对租赁点进行聚类。(3)本文提出了一个基于时间序列的公共自行车租赁点自行车数目预测模型。在现有公共自行车系统租赁点自行车预测模型的基础上,本文采用基于租赁点聚类方法的数据选择方式,并结合租赁点历史趋势,对公共自行车系统租赁点中的自行车数目进行预测。通过实际数据的实验结果与现有模型的预测结果进行对比,本文的预测模型具有较高的准确性。(4)针对预测模型在数据选择过程中,所选择相似的历史数据因影响因素存在差异而产生的误差,本文提出了公共自行车预测结果的误差补偿方法。通过对可能产生误差的因素进行分析,并对这些因素的具体影响大小进行量化,结合在预测模型中使用到的历史数据,运用误差补偿方法计算出存在的误差值,并补偿到预测模型的结果中。实验表明,经过误差补偿以后的预测结果精确度更高。
[Abstract]:The public bicycle system can effectively solve the "last kilometer" problem in people's travel, and improve the overall service level of urban public transport.However, it is one of the main problems of public bicycle system that it is difficult to borrow or return a car, which directly affects the satisfaction of users.Bicycle scheduling is one of the effective methods to solve these problems, and the prediction of bicycle number at rental point is one of the core problems of bicycle scheduling.Therefore, according to the characteristics of public bicycle system, this paper studies the prediction of bicycle number at rental point, which has important theoretical significance and application value.This paper presents a new prediction framework for the number of bicycles at rental points.The framework includes three parts: data selection, prediction model and error compensation. Based on the prediction model and error compensation mechanism, it can greatly improve the accuracy of prediction.In order to provide good data support for the prediction model, this paper proposes a clustering method based on the scale of bicycle use for public bicycle rental points.Through the analysis of different weather (sunny, cloudy, heavy rain, Greater Snow, fog, strong wind, etc.), seasons, date types and different rental points, combined with the actual operation data of public bicycle system in Lucheng District of Wenzhou City,The variation curves of various types of lease points under different external conditions are described and analyzed, and the law and amplitude of change are expressed as characteristic strings.This paper presents a time series based prediction model for the number of bicycles in public bicycle rental points.On the basis of the existing prediction model of bicycle rental point in public bicycle system, this paper adopts the data selection method based on rent-point clustering method, and combines the historical trend of rental point.The number of bicycles in the rental point of the public bicycle system is predicted.By comparing the experimental results of the actual data with the prediction results of the existing models, the prediction model in this paper has a high accuracy.This paper presents an error compensation method for the prediction result of public bicycle, which is caused by the difference of influencing factors in the similar historical data.By analyzing the factors that may produce errors and quantifying the specific influence of these factors, combining with the historical data used in the prediction model, the error compensation method is used to calculate the error.And compensation to the results of the prediction model.The experimental results show that the accuracy of the prediction results after error compensation is higher.
【学位授予单位】：南京师范大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.225

【参考文献】