SEI+II Information Integration through Events

许多环境观测是在不同的空间和时间尺度上收集的，这排除了数据的容易整合，并阻碍了对生态系统动力学的更广泛了解。海洋观测系统提供了一个多传感器系统在不同时空系统中观测几个变量的具体例子。该项目基于将不同的时空环境传感器流的信息内容转换为一个共同的更高级别的抽象：时空事件数据类型，来集成不同的时空环境传感器流。时空事件数据类型跨观测级别数据的多样性进行标准化，以产生用于探测和分析的通用数据类型。缅因湾海洋观测系统(GOMOOS)的数据提供了多变量的时间和空间-时间序列，空间-时间事件从这些序列中被探测和组合。事件检测采用自上而下和自下而上相结合的方法。自上而下组件指定事件本体，而自下而上组件基于从时间和空间-时间序列中提取原始事件(例如，减少、增加、局部最大值和最小值序列)。对提取的事件的探测和分析采用了基于称为事件带的图形基元的图形探索环境，并将其组成事件带堆栈和面板，以支持对各种时空模式的调查。该项目基于事件的概念，提出了一种新的信息集成方法，可扩展到社会经济、金融、立法、监测和卫生相关信息等多个领域。该项目将为时间和空间-时间序列中的事件探测贡献新的数据挖掘战略，并为审查和发展关于时空事件模式和相互作用的假设提供一套灵活的探索性工具。

Many environmental observations are collected at different space and time scales that preclude easy integration of the data and hinder broader understanding of ecosystem dynamics. Ocean Observing Systems provide a specific example of multi-sensor systems observing several variables in different space - time regimes. This project integrates diverse space-time environmental sensor streams based on the conversion of their information content to a common higher-level abstraction: a space-time event data type. The space-time event data type normalizes across the diversity of observation level data to produce a common data type for exploration and analysis. Gulf of Maine Ocean Observing System (GOMOOS) data provide the multivariate time and space-time series from which space-time events are detected and assembled. Event detection employs a combined top down-bottom up approach. The top down component specifies an event ontology while the bottom up component is based on extraction of primitive events (e.g. decreasing, increasing, local maxima and minima sequences) from time and space-time series. Exploration and analysis of the extracted events employs a graphic exploratory environment based on a graphic primitive called an event band and its composition into event band stacks and panels that support investigation of various space-time patterns.The project contributes a new information integration approach based on the concept of an event that can be extended to many domains including socio-economic, financial, legislative, surveillance and health related information. The project will contribute new data mining strategies for event detection in time and space-time series and a set of flexible exploratory tools for examination and development of hypotheses on space-time event patterns and interactions.