Stochastic comparison of machine learning approaches to calibration of mobile air quality monitors

Recently, the interest in the development of new pervasive or mobile implementations of air quality multisensor devices has significantly grown. New application opportunities appeared together with new challenges due to limitations in dealing with rapid pollutants concentrations transients both for static and mobile deployments. Sensors dynamic is one of the primary factor in limiting the capability of the device of estimating true concentration when it is rapidly changing. Researchers have proposed several approaches to these issues but none have been tested in real conditions. Furthermore, no performance comparison is currently available. In this contribution, we propose and compare different approaches to the calibration problem of novel fast air quality multisensing devices, using two datasets recorded in field. Machine learning architectures have been designed, optimized and tested in order to tackle the cross sensitivities issues and sensors inherent dynamic limitations to perform accurate prediction and uncertainty estimation. Comparison results shows the advantage of dynamic non linear architectures versus static linear ones with support vector regressors scoring best results. © Springer International Publishing AG 2018.