关键词：医疗数据；医药；上下文
摘 要：Medical data are intrinsically context-dependent, and cannot be properly interpreted outside of their specific contexts. Therefore, data analysis, especially, secondary data analysis, such as data mining, must incorporate contextual information. This chapter discusses the need for an explicit context representation in medical data mining. It focuses on five contextual dimensions: goal orientation, interdependency of data, time sensitivity, source validity, and absent value semantics. It demonstrates context-dependent modeling based on examples of clinical data used for screening, diagnosis, and research of a serious respiratory disorder, obstructive sleep apnea (OSA). In particular, the chapter describes context-dependent interpretation for three OSA risk factors: large neck circumference, snoring, and smoking. Furthermore, it presents a conceptual framework for representation of the contextual information. This framework is based on a semiotic approach to represent multiple interpretations of data and a fuzzy-logic approach to represent vagueness of data.