Comparison of Self Organizing Maps Clustering with Supervised Classification for Air Pollution Data Sets

Abstract

Air pollution is a serious problem of modern urban centers. The objective of this research is to investigate the problem by using Machine Learning techniques. It comprises of two parts. Firstly, it applies a well established Unsupervised Machine Learning approach (UML) namely Self Organizing Maps (SOM) for the clustering of Attica air quality big data vectors. This is done by using the concentrations of air pollutants (specific for each area) for a period of 13-years (2000-2012). Secondly, it employs a Supervised Machine Learning methodology (SML) by using multi layer Artificial Neural Networks (ML-ANN) to classify the same cases. Actually, the ANN models are used to evaluate the SOM reliability. This is done, because there is no actual and well accepted clustering of the related data to compare with the outcome of the SOM and this adds innovation merit to this paper.