Efficiently Labeling and Retrieving Temporal Anomalies in Relational Databases

Time and temporal constraints are implicit in most databases. To facilitate data analysis and quality assessment, a database should provide explicit operations to identify the violation of temporal constraints. Against this background, the purpose of this paper is threefold: (1) we identify and provide a formal definition of five common anomalies in temporal databases, (2) we propose two new relational operations that allow, respectively, to label anomalous tuples in and to retrieve the anomalous tuples from a dataset, and (3) we provide three different SQL implementations of these operations for current relational database management systems. The healthcare domain is used to illustrate the usage and utility of the temporal anomalies. Finally, an experimental evaluation on real-world and synthetic data analyses the performance of the different implementations of the anomaly operators.