A Real-Time Atypical Amplitude Detection based on Measurement Metadata
Mario José Diván · 2019
PAbMM is a data stream engine which implements a real-time data processing strategy focused on the Measurement and Evaluation (M&E) projects. This allows an active monitoring of different kinds of the entities under analysis specified in the projects. The M&E projects are defined using an M&E framework with the aim of keeping the comparability and extensibility of the projects, fostering the comparability of the measures. The processing strategy runs a set of the online statistical analysis for real-time detecting inconsistencies in relation to the project definition (e.g. a data source miscalibration). In this work, a new component oriented to detect an atypical amplitude on the data in real-time using the measurement metadata related to the project definition is introduced. The underlying idea related to the new component and its incorporation in the processing strategy is shown. In addition, a simulation on the monitoring of the heterogeneous data sources is shown, which gives a processing rate of 2000 measures by each 0,7 seconds. It incorporates a complement to the current outlier detection strategy because the variation range could be atypical, even when all the values are considered inside the expected values.