基于多元线性回归模型的缺失浮动车数据填充研究

发布时间：2018-05-09 21:08

本文选题：浮动车数据 + 多元线性回归模型　；参考：《哈尔滨工业大学》2015年硕士论文

【摘要】：在现实生活中,数据缺失问题是很广泛存在的,无论是在交通方面还是在社会经济研究、生物医药研究等诸多领域中数据缺失现象都是不可避免的。因为数据存在缺失,不单会增加分析研究任务的复杂程度,这样既会大大降低了统计工作的效率,又会导致统计分析结果的重大偏差。所以,为了得到较为完整的数据,采用数理统计的方法对缺失的数据进行填充,是数据处理中不可缺少的重要步骤。本文就是以浮动车数据为例,来研究缺失数据的填充方法。本文研究的主要内容是,将深圳市路网与浮动车数据相结合,得到路网中存在的缺失数据,为了填充缺失部分提出多元线性回归模型,尽可能使得数据覆盖路网范围更广,形成路况发布指南,方便人民出行。具体如下:考虑到交通数据的时空相关性,分析在多尺度下路网的空间相关性,得到缺失数据插补的空间相关因素,同时分析浮动车数据的时间相关性,确定了时间窗的尺度,为后文插补缺失数据模型奠定基础。结合时空相关性,应用多元线性回归模型。首先仅结合空间相关性建立模型,通过选取训练数据做验证分析,效果不好,精度较低;为了提高精度引入时间相关性因素建立模型,进行对比验证,得到在结合时空关系的多元线性回归模型填充缺失数据更具有普遍适用性,并总结该模型适用的四种情况,同时根据课题组成员针对热点区域的研究得到的三个热点区域,分别进行遍历填充。最后是实证分析部分。本文通过对热点区域福田区为例,选取训练数据对模型进行实证校验,通过实证数据校正模型的准确性,然后对实际道路缺失的数据进行填充并与该缺失部分历史存在数据做佐证,进行路况发布。本文的研究能够得到一个结合时空相关性填充缺失数据的可靠模型。
[Abstract]:In real life, the problem of missing data is very widespread, whether in the transportation or in the social and economic research, biomedical research and many other fields of data missing phenomenon is inevitable. The lack of data not only increases the complexity of the task of analysis and research, but also greatly reduces the efficiency of statistical work and leads to a significant deviation of the results of statistical analysis. Therefore, in order to obtain more complete data, it is an indispensable and important step in data processing to use mathematical statistics to fill the missing data. This paper takes floating car data as an example to study the filling method of missing data. The main content of this paper is to combine the data of Shenzhen road network and floating car to get the missing data in the road network. In order to fill the missing part, a multivariate linear regression model is proposed to make the data cover the road network more widely. Form road condition issue guide, convenient people travel. The details are as follows: considering the temporal and spatial correlation of traffic data, the spatial correlation of road network under multi-scale is analyzed, and the spatial correlation factors of missing data interpolation are obtained. At the same time, the temporal correlation of floating vehicle data is analyzed, and the scale of time window is determined. It lays the foundation for the later interpolation missing data model. Combined with temporal and spatial correlation, multiple linear regression model was applied. First of all, only combined with spatial correlation to establish a model, through the selection of training data for verification analysis, the effect is not good, the accuracy is low; in order to improve the accuracy of the introduction of time correlation factors to establish a model, to compare and verify, It is more applicable to obtain the missing data in the multivariate linear regression model combined with space-time relationship, and to summarize the four cases of the model. At the same time, according to the research of the hot spot region by the members of the research group, three hot spots can be obtained. Traversal padding is carried out respectively. The last part is empirical analysis. Through the example of Futian district, the training data is selected to verify the model, and the veracity of the model is corrected by the empirical data. Then the missing data of the actual road are filled and verified with the missing part of the historical data, and the road condition is released. In this paper, we can obtain a reliable model combining spatiotemporal correlation with missing data.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491

【参考文献】