通过分析区域内车辆加油的大数据, 研究车辆加油的普遍行为模式, 调查可能的异常行为. 为此, 以覆盖新疆维吾尔自治区的车辆加油数据为基础, 设计了一个交互式可视分析系统. 首先通过抽取相关数据集的基础特征,得出加油站、汽车、驾驶员3类实体之间的关系; 然后使用多种可视化经典视图并加以组合; 此外, 在部分视图上增加了一些额外的图形元素, 以在具体应用场景下从不同视角描绘出典型的数据特征, 如时空特征等, 同时展示不同实体之间的关系. 通过2个基于真实数据的案例, 在领域专家的协助下分析数据中的典型个体行为模式及统计群组特征, 最终实现对异常行为的识别.%The goal of this research is to study the general vehicle refueling behavior and investigate poten-tial abnormal events through the analysis on massive gas refueling data within a region. Under this context, we design an interactive visualization analysis system based on vehicle refueling data collected from the whole region in Xinjiang province, China. First we extract the basic data features from the dataset and obtain the relationship between entities such as gas stations, vehicles, and drivers. Secondly, we employ multiple classical visualization models and meaningful composition of these models. On several view models, we append additional graphical elements in order to illustrate typical data features (such as geospatial and tem-poral features) from different perspectives under certain real-life application scenarios. In addition, the view models can describe the relations between different entities. Through two real-life case studies, we analyze the typical individual behaviors as well as statistical group features and finally realize the detection of ab-normal refueling events with the assistance of domain experts.
展开▼
