Microblogs are now important platforms for people to communicate their opinions and emotions with others. However, it is difficult to effectively get messages of interest from microblogs, because of their enor mous quantity, dynamics and complexity. In this paper, we present an interactive visual analytics system to ana lyze the data from microblogs. First, an effective SVM classifier is constructed interactively. The SVM classifier and keyword queries are used to filter messages and monitor the topics of interest. Then, we provide several data analysis and visualization tools including statistics tools, heat map, tag cloud. They can assist analysts to explore and understand the data. Additionally, we developed a new illustration method based on video magnification technique to reveal subtle temporal features in the data. The results of our experiment using data from Twitter show that our system can dynamically classify messages according to the topics of interest and reveal tiny pat terns in the data from microblogs.%推特等微博已成为人们交流信息、发表观点的重要平台。针对微博数据海量、动态且复杂多样，导致从微博数据中实时获取有用信息具有一定的难度的问题，设计了一个交互的可视分析系统。系统首先基于关键词过滤和支持向量机分类动态地监控不断获取的微博数据；接着以热力图、统计图、词云等可视化方式呈现微博的过滤效果，以方便用户进行探索；最后借鉴视频运动放大技术对微博数据中的微小时变模式进行放大，能有效地发现数据中的微小特征。以推特数据为例进行实验的结果表明，该系统能动态地对微博数据进行有效分类，并能识别数据中的微小变化模式，帮助用户准确了解事件的发展。
展开▼
