基于移动通信大数据的人口流动性测度研究

发布时间：2018-06-28 21:09

本文选题：流动人口规模 + 移动通信用户数据　；参考：《山西财经大学》2017年硕士论文

【摘要】：人口流动性是经济社会发展的一个重要指标,用来测度人口以流动方式追求经济社会目标而形成的较长时间的自由迁徙和异地生活的状况。依照政府统计口径,流动人口是指在中国户籍制度条件下,离开了户籍所在地到其他地方居住的人口,但目前尚无明确、准确和统一的定义。截止到2016年末,我国流动人口总数约为2.45亿人。经济增长是人口流动的一个重要原因,鉴于流动人口的构成成分复杂、流动周期不确定、流动轨迹多变等原因,我国现有的人口流动性统计存在诸多问题,口径统计不一,数据质量参差,不能满足政府和社会的统计需求,流动人口相关的统计方法和相关制度都亟待改进。本文基于移动通信运营商的即时通话记录数据,以人口行为学特征为基础,从移动通信大数据所表征的用户行为对人口的流动性进行判断和测度,在对流动人口概念进一步界定的基础上,设计了一种将基于机器学习算法构建的流动人口识别模型和基于捕获再捕获抽样构建的人口流动性测度模型相结合来对流动人口规模进行估计的方法。在基于机器学习方法构建流动人口识别模型时,通过对移动通信用户流动人口和本地人口通信行为特征的分析构造了流动人口识别特征变量,利用AUC-RF方法进行了特征变量的选择。在此基础上,本文选择了决策树、Bagging、随机森林、支持向量机以及人工神经网络五种算法进行模型的构建,并通过多种评价标准对模型进行评估和选择,最终选择分类性能和泛化能力最优的随机森林模型作为最终的流动人口识别模型,对样本集中的未分类样本进行了分类预测。在构建基于捕获再捕获抽样的流动人口规模测度模型时,实证表明该估计方法能比较准确可靠的对地区流动人口规模进行估计。因此本文得出结论认为,本文设计的基于移动通信大数据的人口流动性测度方法与传统的流动人口调查方法可并行使用,相互补充,相互印证。本文希望在移动通信大数据的基础上,对改进我国流动人口统计调查探索一个基于大数据思想的统计方法和制度,依托同时期的移动通信记录数据,利用科学的统计推断方法,对流动人口的规模和特征进行估计和外推,从而得到更加精确和完整的人口统计数据。实证检验表明,本文设计的方法成本低、速度快、精度较高,非常适合于对我国现行的统计制度进行改进和拓展。
[Abstract]:Population mobility is an important indicator of economic and social development, which is used to measure the situation of long time free migration and living in different places. According to the government statistics, the floating population refers to the population who left the domicile to live in other places under the condition of Chinese household registration system, but there is no clear, accurate and unified definition at present. By the end of 2016, the total floating population in China was about 245 million. Economic growth is an important reason for population mobility. In view of the complexity of the composition of floating population, the uncertainty of flow cycle and the changeable flow path, there are many problems in the current population mobility statistics in our country. The data quality is uneven and can not meet the statistical needs of the government and society. The relevant statistical methods and systems of floating population need to be improved urgently. Based on the instant call record data of mobile communication operators and based on the characteristics of population behavior, this paper judges and measures the mobility of population from the user behavior represented by mobile communication big data. On the basis of a further definition of the concept of floating population, A new method is designed to estimate the size of floating population by combining the mobile population identification model based on machine learning algorithm and the population mobility measurement model based on capture and recapture sampling. In the process of constructing a mobile population identification model based on machine learning, the mobile population identification feature variables are constructed by analyzing the characteristics of mobile population and local population. The AUC-RF method is used to select the characteristic variables. On this basis, this paper selects five algorithms of decision tree agginging, random forest, support vector machine and artificial neural network to construct the model, and evaluates and selects the model through various evaluation criteria. Finally, a stochastic forest model with optimal classification performance and generalization ability is chosen as the final identification model of floating population, and the unclassified samples in the sample set are classified and forecasted. When constructing the scale measurement model of floating population based on capture and recapture sampling, the empirical results show that this method can estimate the size of floating population accurately and reliably. Therefore, this paper concludes that the method of population mobility measurement based on mobile communication big data can be used in parallel with the traditional method of mobile population survey, which can complement each other and confirm each other. On the basis of mobile communication big data, this paper hopes to explore a statistical method and system based on the idea of big data to improve the survey of mobile population in our country, rely on the mobile communication record data of the same period, and use scientific statistical inference method. The size and characteristics of floating population are estimated and extrapolated to obtain more accurate and complete demographic data. The empirical test shows that the method designed in this paper has the advantages of low cost, fast speed and high precision, which is very suitable for improving and expanding the current statistical system of our country.
【学位授予单位】：山西财经大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：C924.2

【参考文献】