Air Quality Prediction for Multi-Scale Time Period Based on AutoML

Air quality sensors generate vast amounts of underutilized data across diverse fields, including environmental monitoring, agriculture, and data warehousing. Despite the ease of data collection, the value of air quality data remains unclear to many industries, and previous efforts to build custom machine learning models have limited accessibility. These efforts are often inaccessible to a large portion of the industry due to their specificity and resource demands, resulting in valuable data stores being underused. This study introduces an adapted approach using state-of-the-art Automated Machine Learning for time-series IoT sensor data, specifically tailored for air quality monitoring. Our experiments demonstrate that time-series AutoML models achieve comparable or improved accuracy over both traditional and neural network multi-target machine learning methods, including Random Forests, Extra Trees, ElasticNet, and LSTM. These results highlight the feasibility and benefits of accessible machine learning-driven analysis for air quality sensor data, offering cross-industry applicability for those lacking specialized Machine Learning expertise.