Abstract
Data quality is an important issue in almost all application domains. A successful project invariably relies on the high quality data. This thesis proposes methods for analyzing the quality of observational data in real time. An objective of real time quality control is to detect anomalies of observational data. Specifically, this thesis focuses on the detection of a common type of anomaly called outlier. We say a data item is an outlier when its value differs by a specified amount from values of other data items recorded under similar conditions. The thesis presents an overview of outlier detection for real-time quality control, including outlier detection algorithms, local outlier factor (LOF) algorithm and local distance outlier factor (LDOF) algorithm, partially ordered data analysis method, and minimum bounding box method. The thesis also describes several simulation experiments to evaluate the performance of the methods.