Knowledge-based trend detection and diagnosis

摘要: This thesis presents a knowledge-based approach to diagnostic process monitoring. The cornerstone of this work is the representation and detection multivariate trends in data. trend representation, called template, denotes time-varying pattern multiple variables common population. Each contains representations for landmark events set phases, each temporally uncertain. phases are represented by partially ordered temporal intervals. Bound interval constraints on real-valued functions measurable parameters. low-order polynomial regression models, with either qualitative or quantitative coefficient estimates. A computer program TrenDx diagnoses matching data competing templates within clinical context. score template hypothesis based mean absolute percentage error between models not only maintains alternate hypotheses different trends, but also optimizes over chronologies description. Therefore can report both what most significant when take place that trend. The describes how be extended complete an architecture automated faulty judged if time it matches better than expected trend. Significance may trigger alarm, switch context, filter intelligent display. TrenDx has been applied diagnosis two medical domains. pediatric growth from heights, weights, bone ages, sexual staging detects intensive care unit patients hemodynamic respiratory techniques intended as general purpose, applicable other monitoring applications such industrial control, telecommunications, economics, finance. (Copies available exclusively MIT Libraries, Rm. 14-0551, Cambridge, MA 02139-4307. Ph. 617-253-5668; Fax 617-253-1690.)