Use of Abstraction to Simplify Monitoring

ion Tests A careful examination of the heart rate data indicates that most of it has the following form: "8/16/93 23:24:44" 176 "8/16/93 23:26:04" 175 "8/16/93 23:26:24" 176 "8/16/93 23:26:44" 175 "8/16/93 23:27:02" 176 In other words, most of the changes from one reading to the next were insignificant. A more rigorous analysis of the heart rate data indicates that there were no two consecutive values that were the same, but that only 60 out of 627 samples changed by more than one from the previous value. Only 30 samples showed a sample-to-sample change of 5 or more. It is also the case that only gross changes in heart rate are clinically meaningful. If we can introduce such a distinction early in the data processing, then we can save subsequent reasoning tasks the trouble of determining that the information is about the same. To that end we introduce a value abstraction transformation. The following table shows three different abstraction functions that were used in the experiments. The first two break the numeric scale into ranges of 5 and 10 and only use the range of the answer as the persistent value. If two values occur that are in the same range, then that is treated as no change in the data value. (Any temporal persistence is still used, though). “Stratify” divides the numeric range into three categories. Heart Rate Abstract 5 Abstract 10 Stratify