Analyzing and Mining Large Uncertain Data Sets: models, methods and applications

在一些重要的应用中，不确定数据是固有的，例如使用传感器网络的环境监测、使用客户调查的市场分析和量化经济学研究。这些应用中的不确定数据通常是由数据随机性和不完整性、测量设备的限制、数据更新延迟等因素造成的。例如，传感器经常用于检测偏远或保护区内是否存在濒危、受威胁或特别关注的风险类别的动物。由于传感器的局限性，检测并不总是准确的。取而代之的是，通常通过估计检测置信度来模拟不确定性。分析和挖掘从这些传感器收集的不确定数据可以使生物学家捕捉和理解野生动物有趣的行为模式。

Uncertain data are inherent in some important applications, such as environmental surveillance using sensor netoworks, market analysis using customer surveys, and quantitative economics research. Uncertain data in those applications are generally caused by factors like data randomness and incompleteness, limitations of measuring equipment, delayed data updates, etc. For example, sensors are often used to detect presence of endangered, threatened, or special concern risk categories of animals in remote or preserved regions. Due to limitations of sensors, detections cannot be accurate all the time. Instead, detection confidence is often estimated to model uncertainty. Analyzing and mining uncertain data collected from such sensors can enable biologists to capture and understand interesting behavior patterns of wild animals.