A knowledge-based framework for the management of imprecision in data

不精确性是许多数据固有的，特别是在复杂的生物、生态和医学系统中。通常，现实生活中的决策是基于主观的，不确定的和不精确的数据。因此，不精确性的管理是数据共享、决策制定、信息披露和数据挖掘中最关键的问题之一。不精确性已经在几个学科的背景下进行了研究，并且已经开发了各种方法来表示数据库中的不精确数据，最近，在Web存储库中。然而，目前的方法并没有充分解决领域知识的背景下的不精确性的解释。然而，对不精确的解释是高度语境化的。例如，"血压高于正常"的陈述在一种情况下可能足够精确，或者在另一种情况下可能不够精确，并且可能需要诸如"160/90"的测量和"没有抗高血压药物"的附加信息。 拟议的研究项目将回答三个问题：（1）如何对数据中不同类型和水平的不精确性进行建模，（2）如何整合具有不同不精确性水平的数据，以及（3）如何对可接受和不可接受的不精确性概念进行建模和管理。该项目将创建一个实用的知识框架和软件工具，以表示和管理数据中各种形式的不精确性。该框架和工具将在两个复杂领域的背景下进行评估：医疗决策和生物数据分类。评估将基于两个真实案例：（1）用于诊断和治疗睡眠障碍的医疗数据仓库的构建和（2）用于分类和分析槲寄生植物的植物学数据仓库的构建。研究结果将适用于心理学、认知科学和生态学等知识密集型领域的数据表示和处理。

Imprecision is intrinsic to many data, especially in complex biological, ecological, and medical systems. Often, real-life decisions are made based on subjective, uncertain, and imprecise data. Thus, the management of imprecision is one of the most critical issues in data sharing, decision making, information disclosure, and data mining. Imprecision has been studied in the context of several disciplines and various methods have been developed to represent imprecise data in databases, and, recently, in the Web repositories. However, the current approaches do not address sufficiently the interpretation of imprecision in the context of domain knowledge. Yet, the interpretation of imprecision is highly contextual. For example, a statement "blood pressure above normal" may be sufficiently precise in one context or it could be insufficient in another context and a measurement such as "160/90" and additional information "no antihypertensive medication" may be needed. The proposed research project will answer three questions: (1) how to model different types and levels of imprecision in data, (2) how to integrate data characterized by various levels of imprecision, and (3) how to model and manage the notion of acceptable and unacceptable imprecision. The project will create a practical knowledge-based framework and software tools to represent and manage various forms of imprecision in data. The framework and tools will be evaluated in the context of two complex domains: medical decision-making and biological data classification. The evaluation will be based on two real-life cases: (1) construction of a medical data warehouse for the diagnosis and treatment of sleep disorders and (2) construction of a botanical data warehouse for the classification and analysis of mistletoe plants. The results will be applicable to data representation and processing in other knowledge-intensive domains, such as psychology, cognitive science, and ecology.