This study investigates the use of Long Short-Term Memory (LSTM) neural networks, Bayesian Optimization, and Bayesian Optimization HyperBand (BOHB) for calibrating low-cost air quality sensors monitoring PM2.5 concentrations in Region TBA. These models are evaluated for their ability to improve sensor accuracy under the region’s unique environmental conditions, including temperature fluctuations and high humidity. LSTM, known for its capability in handling time-series data, is compared with Bayesian Optimization, which tunes model hyperparameters to achieve optimal performance, and BOHB, an advanced technique that balances exploration and exploitation for more efficient hyperparameter tuning. The comparison focuses on calibration accuracy, computational efficiency, and suitability for real-time applications. Findings will inform the development of more precise and scalable methods for air quality monitoring in diverse environments.
Application of LSTM, Bayesian Optimization, and Bayesian Optimization HyperBand for Calibrating Low-Cost Air Quality Sensors in Region TBA
Category
Poster Abstract
Description
Submitted by:
Jennifer Smith George Mason University - Computational Sci. - Geoinfo. Sciences jsmit92@gmu.edu