出租车需求量预测模型的研究

发布时间：2018-06-24 08:37

本文选题：多元线性回归模型 + 随机森林回归模型　；参考：《大连海事大学》2017年硕士论文

【摘要】：由于社会经济的快速发展,出租车逐渐成为了人们出行的首选方式。近年来随着网络技术的发展和智能终端的普及,网络打车平台下的出租车(以下简称出租车)凭借其快速、便捷和优质的服务逐渐的成为人们出行的首选。然而,在现实中出租车司机很难知道城市中不同时刻不同区域的出租车需求量,这样可能会导致出租车出现空载或者供不应求的现象,极大的浪费了社会资源。为了解决这一问题本文研究出租车需求量预测模型,对未来不同时刻的出租车需求量进行预测。本文首先研究了出租车需求量预测问题的常用研究方法以及相关的短时交通流预测模型,对常用模型进行了分析指出了各自的特点。通过使用数据可视化的方式验证了天气状况、PM2.5、温度和交通拥堵状况等因素对出租车的短时需求具有一定的影响,然后借鉴短时交通流预测问题构造变量的方式对样本的变量进行了提取和设计,为模型的建立奠定了基础。接着本文构建了多元线性回归模型、随机森林回归模型、梯度渐进回归树模型,并基于这三个模型提出了一个线性变权重组合预测模型来预测出租车的需求量。其中组合预测模型的权重是根据单一模型各自的历史预测误差的均方倒数来确定的,并且这个模型的权重是实时调整的。最后本文以中国某一线城市出租车需求量最大的区域的数据为例,验证本文构建的组合预测模型的有效性,实验表明线性变权重的组合预测模型的整体预测精度要高于单一预测模型的预测精度。为了保证模型的通用性和便捷性,本文利用Rserve实现了 JAVA语言与R语言交互的出租车需求量预测系统,考虑到系统的响应时间,该系统采用C/S架构。
[Abstract]:With the rapid development of social economy, taxi has gradually become the preferred way for people to travel. In recent years, with the development of network technology and the popularization of intelligent terminals, taxi (taxi) under the network taxi platform has gradually become the first choice for people to travel by virtue of its fast, convenient and high quality service. However, in reality, it is difficult for taxi drivers to know the demand for taxis in different areas at different times in the city, which may lead to the phenomenon of no load or short supply of taxis, which is a great waste of social resources. In order to solve this problem, this paper studies the forecast model of taxi demand and forecasts taxi demand at different times in the future. This paper first studies the common research methods of taxi demand forecasting and the related short-term traffic flow forecasting models, and points out the characteristics of the common models. Through the use of data visualization to verify that weather conditions such as PM2.5, temperature and traffic congestion, and other factors have a certain impact on the demand for taxis in the short term. Then the sample variables are extracted and designed by using the method of constructing variables for short time traffic flow forecasting problem, which lays a foundation for the establishment of the model. Then, the multivariate linear regression model, the stochastic forest regression model and the gradient progressive regression tree model are constructed. Based on these three models, a linear variable weight combination forecasting model is proposed to forecast the taxi demand. The weight of the combined prediction model is determined according to the mean square reciprocal of the historical prediction error of the single model, and the weight of the model is adjusted in real time. Finally, taking the data of the regions with the largest taxi demand in a first-tier city in China as an example, the validity of the combined forecasting model is verified. The experimental results show that the overall prediction accuracy of the combined forecasting model with linear variable weights is higher than that of the single prediction model. In order to ensure the generality and convenience of the model, a taxi demand forecasting system based on Java language and R language is implemented by Rserve. Considering the response time of the system, the system adopts C / S architecture.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O212.4

【参考文献】