基于社交媒体的人类移动时空规律研究

发布时间：2018-03-06 12:18

本文选题：社交媒体　切入点：数据挖掘　出处：《武汉大学》2017年硕士论文　论文类型：学位论文

【摘要】：人类的活动影响着交通、住房、商业、文化、基础设施建设等与城市发展相关的方方面面,认识并了解人类的活动有助于城市的规划与建设。近年来,基于位置的社交媒体平台的发展,使人们的生活从现实世界延伸到虚拟世界中,人们在现实世界中活动时在部分时间里也同时活动于虚拟世界里,位置和时间是联系两个世界的桥梁,人们在虚拟世界中活动时留下了在现实世界所处的位置和时间信息。我们可以根据用户在虚拟世界中留下时空信息研究人类在现实世界中的移动规律。现实生活中的绝大多数人都有固定的生活节奏,因此其活动都有一定的规律可循,然而由于生活的复杂性,人们在遵循规律性活动的同时也进行了一些偶然性活动。如何从偶然活动中提取规律性活动成为研究人类移动规律的一个挑战。本文基于社交媒体数据,使用扩展后的Markov模型研究了人类各类活动所占比重,并分析了群体的活动对城市人口流动的反映。之后使用时空路径理论提取了人类主要的活动模式,并使用聚类算法根据活动模式将用户划分到不同类别,研究了某些类别具有的时空特征。本文所做主要工作如下:1)基于Markov模型中状态转移思想,将时间维度加入到模型中来,研究人类在不同时段出现在不同位置以及在位置间移动的可能性,包括:人类移动提取,人类移动位置出现探测、人类移动位置转换探测,综合预测算法设计。2)将个体移动规律的探测方法应用到群体移动规律探测中,使用活动位置出现概率反映城市人群在不同时刻的聚集状况,使用活动位置转换概率反映人群在不同时刻的流动情况。并使用ECharts动态展示人群的动态活动情况。3)将时空路径理论应用到人类的长期的主要活动模式探索中来,研究人类移动热点的提取及聚类方法,时空路径生成及路径出现概率计算。使用参数组合生成了用户多条时空路径,获得用户更全面的活动模式。将时空路径在以24小时为Z轴的三维空间中展示以获得人类于一天内的移动规律及移动规律随时间的变化。4)不同用户的时空路径具有不同的时空特征,代表了不同的活动模式。设计时空路径聚类方法,将不同时空特征的时空路径划分到不同类别,研究不同类别用户具有的时间和空间上的规律。
[Abstract]:Human activities affect transportation, housing, commerce, culture, infrastructure construction and other related aspects of urban development, understanding and understanding of human activities contribute to urban planning and construction in recent years, With the development of location-based social media platform, people's life extends from the real world to the virtual world. Location and time are bridges between the two worlds, When people move in virtual world, they leave the information of their position and time in the real world. We can study the law of human movement in the real world according to the time and space information left by the user in the virtual world. Most of them have a fixed rhythm of life, So they all have certain rules to follow, but because of the complexity of life, While following regular activities, people have also carried out some accidental activities. How to extract regular activities from accidental activities has become a challenge in studying the laws of human mobility. This paper is based on social media data. The proportion of various human activities is studied by using the extended Markov model, and the reflection of group activities on urban population flow is analyzed. Then, the main human activity patterns are extracted by using space-time path theory. We use clustering algorithm to divide users into different categories according to their activity patterns, and study the space-time characteristics of some categories. The main work of this paper is as follows: 1) based on the idea of state transition in Markov model, the time dimension is added to the model. To study the possibility that human beings appear in different positions at different times and move between positions, including: human movement extraction, human mobile position detection, human mobile position conversion detection, The synthetic prediction algorithm design. 2) apply the detection method of individual movement law to the detection of group movement law, and use the probability of occurrence of activity position to reflect the gathering state of city crowd at different time. The temporal and spatial path theory is applied to the exploration of human's long term main activity pattern, using the probability of changing the activity position to reflect the movement of the crowd at different times, and using ECharts to show the dynamic activity of the crowd dynamically. The extraction and clustering methods of human mobile hot spots, the generation of space-time paths and the calculation of path occurrence probability are studied, and the multiple spatio-temporal paths of users are generated by using the combination of parameters. Obtain a more comprehensive user mode of activity. Display the space-time path in a three-dimensional space with the Z axis of 24 hours to obtain the movement laws of human beings within a day and their movement laws over time. 4) the space-time paths of different users. Have different space-time characteristics, The method of spatio-temporal path clustering is designed to divide the spatio-temporal paths of different spatio-temporal characteristics into different categories and to study the temporal and spatial laws of different types of users.
【学位授予单位】：武汉大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：C912.1;P208

【相似文献】