【作者】 金玲；

【导师】 王章野； 巫英才；

【作者基本信息】 浙江大学 ， 计算机技术， 2016， 硕士

【摘要】 传感器技术的发展,使得人们能够收集到多种多样的城市数据,如气象数据领域的气象数据、环境领域的空气质量数据、交通领域的交通流数据等。城市数据大多蕴含丰富的城市信息,如果能够有效利用,将极大地帮助我们认识和解决城市所面临的各种问题。其中,一个重要的思路是从多个来源的异构城市数据中挖掘出跨领域的关联模式。通过有效地分析这些跨领域的关联模式,能够服务于多种应用,如空气质量诊断、商业地址的选取等。然而,研究人员在分析、理解、检查计算出的关联模式的过程中,会遇到如下的挑战：1)关联模式的数量很大。成千上万条的关联模式,使得研究人员解读起来非常困难。2)数据结构复杂多样。首先,数据是关联型的数据,因此包括多个维度：其次,各关联项的值有区间值(数值型观测项)和整型数值(类别型观测项)两种,有非研究目标和研究目标之分。另外,数据还包括时空属性和概率属性。为此,本文提出了一套综合的可视分析系统,帮助用户直观地分析和理解关联模式数据。文章结合可视化和交互技术,设计了多个功能模块,使用户方便地从不用的角度、在不同的细节层面上对关联模式数据进行分析和理解。另外,文章结合多视图关联的方法,使各个功能模块紧密结合,帮助用户全面地理解数据。最后,文章通过两个案例证明了本文的系统能够帮助用户直观有效地分析和理解上述复杂的关联模式数据。更多还原

【Abstract】 With the development of sensing technology, people can collect a variety of urban data, such as meteorology data in the meteorology domain, air quality data in the environment domain, and traffic flow data in the traffic domain. We can understand the problems in cities better by properly processing the urban data, since it implies rich knowledge about cities. Among this, a meaningful idea is extracting correlation patterns from cross domain urban data. Such cross domain correlation patterns can serve many applications, such as air quality diagnosis, selection of business addresses. However, the researchers encountered the following challenges in analyzing, understanding and examining the correlation patterns:1. It contains a large number of correlation patterns. Thousands of correlation patterns.2. The data structure of the correlation patterns is very complex. First, it contains multiple dimensions since it implies some correlation. Second, the value of each dimension is either a range(for numerical observation) or a category(for categorical observation).In addition, it contains three other properties, including spatial property, temporal property and probability property.In this paper, we proposed a visual analysis system to help analysts visually analyze and make sense of the correlation patterns. With an integration of multiple visualization and interaction methods, we design a few linked visualizations to allow users to understand and analyze the data from different views on different levels of detail. Through the linked visualizations, the system provides analysts with a quick and comprehensive overview of the data. In the end, case studies are conducted to demonstrate the effectiveness of our system in understanding and analyzing the complex correlation data.更多还原