Real-Time AQI Forecasting With High-Throughput Using IoT Devices

This work explores the real-time and high-throughput decision-making of the Air Quality Index (AQI) and the pollutants concerning commercially available IoT-based de-vices such as GPU and Raspberry Pi. The proposed methodology trains various parameterized algorithms to investigate the best suitable settings that could effectively forecast the future values of AQI, and concerning pollutants. The algorithm with the best performance has been further deployed on different commercially available computing devices for analyzing the throughput. The employed experimental set-up successfully evaluates the best suitable algorithm i.e. Random Forest (RF), with a coefficient of determination (Adj. R2) 0.82. Additionally, deploying the RF algorithm on coral TPU-embedded Raspberry Pi 4 provides 3–11 fold relative enhancement in throughput comparison to CPU, GPU, and Raspberry Pi 4.