Demonstration of an application for remote outlier detection and accommodation

This work describes a tool for data analysis and time series cleaning, with a specific focus on outlier detection and accommodation. Data can be previously collected or obtained in real-time from a remote laboratory experiment using a wireless sensor network (WSN). The tool, presented as a local Matlab Graphical User Interface or as a remote application, allows the user to fill in missing data, resample and select a wide range of state-of-the-art and classical outlier detection and accommodation methods such as Kernel PCA, Modified Z-Score and Grubb´s Test. After the accommodation, the different methods can be compared one with another, as well as with the original series, providing several scalar metrics, such as Euclidian Distance or Complex-Invariant Time Distance, and graphical comparison metrics. Thus, this tool will provide quick access to time series cleaning techniques, applicable, in particular, to data from industrial or clinical scenarios.