Study on Reconstruction of Monitoring Information
Zhaohua Lin · Jisuanji fangzhen · 2008
Model-based methods are used for data analysis and detection abnormal status in monitoring information system. Often, these models are obtained by fitting convenient model structures to observed data. Real measurement data records frequently contain outliers, which can badly degrade the results of an empirical model identification procedure. Analysis methods for outliers were proposed. First, the outliers were located based on Hampel identifier; and second, the methods based on kalman filter were used to compute the predictions correctly for dealing with outliers. Some real datasets including heart rate, central venous pressure, and oxygen saturation of blood from PhysioNet were used for test. The results show that the methods described are often extremely effective in practice for outliers’ analysis for monitoring information system.