Metropolitan air pollution is a growing concern in both developing and developed countries. Fixed-station monitors, typically operated by governments, offer accurate but sparse data, and are increasingly being augmented by lower fidelity but denser measurements taken by mobile sensors carried by concerned citizens and researchers. In this paper, we introduce HazeEst-a machine learning model that combines sparse fixed-station data with dense mobile sensor data to estimate the air pollution surface for any given hour on any given day in Sydney. We assess our system using seven regression models and tenfold cross validation. The results show that estimation accuracy of support vector regression (SVR) is similar to decision tree regression and random forest regression, and higher than extreme gradient boosting, multi-layer perceptrons, linear regression, and adaptive boosting regression. The air pollution estimates from our models are validated via field trials, and results show that SVR not only yields high spatial resolution estimates that correspond well with the pollution surface obtained from fixed and mobile sensor monitoring systems, but also indicates boundaries of polluted area better than other regression models. Our results can be visualized using a Web-based application customized for metropolitan Sydney. We believe that the continuous estimates provided by our system can better inform air pollution exposure and its impact on human health.

This paper introduces a novel concept for creating an ensemble of classifiers. The concept is based on generating an ensemble of classifiers through clustering of data at multiple layers. The ensemble classifier model generates a set of alternative clustering of a dataset at different layers by randomly initializing the clustering parameters and trains a set of base classifiers on the patterns at different clusters in different layers. A test pattern is classified by first finding the appropriate cluster at each layer and then using the corresponding base classifier. The decisions obtained at different layers are fused into a final verdict using majority voting. As the base classifiers are trained on overlapping patterns at different layers, the proposed approach achieves diversity among the individual classifiers. Identification of difficult-to-classify patterns through clustering as well as achievement of diversity through layering leads to better classification results as evidenced from the experimental results.