Abstract
Predicting the status of particulate air pollution is extremely important in terms of preventing possible vascular and lung diseases, improving people’s quality of life and, of course, actively counteracting pollution magnification. Hence, there is great interest in developing methods for pollution prediction. In recent years, the importance of methods based on classical and more advanced neural networks is increasing. However, it is not so simple to determine a good and universal method due to the complexity and multiplicity of measurement data. This paper presents an approach based on Deep Learning networks, which does not use Bayesian sub-predictors. These sub-predictors are used to marginalize the importance of some data part from multisensory platforms. In other words—to filter out noise and mismeasurements before the actual processing with neural networks. The presented results shows the applied data feature extraction method, which is embedded in the proposed algorithm, allows for such feature clustering. It allows for more effective prediction of future air pollution levels (accuracy—92.13%). The prediction results shows that, besides using standard measurements of temperature, humidity, wind parameters and illumination, it is possible to improve the performance of the predictor by including the measurement of traffic noise (Accuracy—94.61%).
Highlights
The subject of air quality study and prediction is a very important research area [1–3].As reported by the authors [4], air pollution prediction methods can be divided into statistical, numerical, neural network and hybrid models
This paper presents experimental results for a neural prediction system for particulate air pollution
Various research centers are working on improving this mechanism, but this work presents a different approach. This approach is based on not using marginalization, but direct inference from raw tabular data, transformed to image form using a special feature extractor. This approach has improved the quality of prediction and in the future can be applied for other branches of technology than air pollution prediction, such as predictive maintenance, medical, economic and social sciences
Summary
The subject of air quality study and prediction is a very important research area [1–3]. The novelty of the paper is the use of Deep Learning algorithm, in addition to classical Neural Network architectures, to extract the features of the raw tabular data vector. This approach allows the attainment of better results for the prediction of future air pollution. Various research centers are working on improving this mechanism, but this work presents a different approach This approach is based on not using marginalization, but direct inference from raw tabular data, transformed to image form using a special feature extractor. This approach has improved the quality of prediction and in the future can be applied for other branches of technology than air pollution prediction, such as predictive maintenance, medical, economic and social sciences
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
