Volterra Graph-Based Outlier Detection for Air Pollution Sensor Networks

Today's air pollution sensor networks pose new challenges given their heterogeneity of low-cost sensors and high-cost instrumentation. Recently, with the advent graph signal processing, network measurements have been successfully represented by graphs depicting relationships between sensors. However, one main problems these is reliability, especially due to inclusion sensors, so detection identification outliers extremely important for maintaining quality data. In order better identify composing a network, we propose Volterra graph-based outlier (VGOD) mechanism, which uses learned from data Volterra-like reconstruction model detect localize abnormal in networks. The proposed unsupervised decision process compared other methods, state-of-the-art methods non-graph-based showing improvements both localization anomalous measurements, that can be corrected malfunctioning replaced.