数据挖掘技术在道路阻抗函数问题中的应用研究

发布时间：2018-08-17 11:31

【摘要】：随着大数据时代的到来,数据仓库、数据挖掘技术发展的非常迅速,利用现有系统中流转和沉淀的数据,挖掘出有用的模式以指导决策,已经成为了大数据时代的趋势。随着突发事件和自然灾害的增多,成品粮应急物流越来越被国家和研究者所重视,为了构建科学、高效、可靠的成品粮应急调度决策系统,本文提出了使用数据挖掘技术解决成品粮应急调度中的配送路径优化问题的研究思路。本文研究了数据挖掘技术中的主要算法和评估指标,对各个算法的优缺点进行了分析和总结。通过将数据挖掘技术与成品粮应急调度决策相结合,针对其中难以解决的动态路径优化问题,提出了使用数据挖掘技术中的回归分析技术进行道路阻抗函数的研究。首先,本文研究从实际数据出发,在收集到北京市2012年1月份环路微波检测道路数据的基础上,采用分布式数据处理手段对该历史道路数据进行了清洗和处理,并借此研究了分布式系统hadoop的数据存储和处理原理,为了方便后续模型研究,本文给出了大规模数据集下特征提取的思路,并设计了道路阻抗函数研究中的特征提取步骤并给出了示例。然后,在模型研究方面,本文利用了数据挖掘技术中的分类、回归技术对输入的特征和目标值进行拟合,以进行道路阻抗函数的确定。在该部分的研究中本文分四个阶段由简单到复杂地对道路阻抗函数进行研究：线性模型研究、基于BPR函数模型的研究、分类回归树模型研究,然后创新性地提出概率性分类回归模型并对其进行了深入研究。以上四个类型的模型,在本文中都给出了详细的公式推导、求解方法以及模型优缺点分析。最后,在模型的验证方面,本文为四个模型分别设计了详细的实验步骤,然后通过对北京市2012年1月的实际历史道路数据进行实验,使用量化的指标对各个模型的实验结果进行对比、分析和验证。通过实验可以证明本文提出的概率性分类回归模型在道路阻抗函数的拟合上表现的最好,最具实用价值。
[Abstract]:With the arrival of the big data era, data warehouse and data mining technology are developing very rapidly. It has become the trend of the big data era to mine useful patterns to guide the decision making by using the data transferred and deposited in the existing system. With the increase of emergencies and natural disasters, the emergency logistics of finished grain has been paid more and more attention by the state and researchers. In order to build a scientific, efficient and reliable decision system for emergency dispatch of finished grain, In this paper, the research idea of using data mining technology to solve the problem of distribution path optimization in emergency dispatch of finished grain is put forward. In this paper, the main algorithms and evaluation indexes of data mining technology are studied, and the advantages and disadvantages of each algorithm are analyzed and summarized. Based on the combination of data mining technology and emergency dispatch decision of finished grain, aiming at the dynamic path optimization problem which is difficult to solve, this paper puts forward the research of road impedance function using regression analysis technology in data mining technology. Firstly, based on the actual data and the data collected from January 2012, this paper uses distributed data processing method to clean and process the historical road data. The principle of data storage and processing in distributed system hadoop is studied. In order to facilitate the subsequent model research, the idea of feature extraction based on large-scale data sets is given in this paper. The feature extraction steps in the study of road impedance function are designed and an example is given. Then, in the research of model, this paper uses the classification of data mining technology, regression technology to fit the input characteristics and target values, in order to determine the road impedance function. In this part of the study, the road impedance function is studied from simple to complex in four stages: linear model, BPR function model, classification regression tree model. Then the probabilistic classification regression model is proposed and studied. In this paper, the formula derivation, the solution method and the advantages and disadvantages of the above four models are given. Finally, in the verification of the model, this paper designs the detailed experimental steps for the four models, and then through the actual historical road data of Beijing in January 2012 to carry out the experiment. The experimental results of each model are compared, analyzed and verified by using quantitative indexes. It can be proved by experiments that the probabilistic classification regression model presented in this paper has the best performance in the fitting of road impedance function and has the most practical value.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491

【参考文献】