Air pollution is one of the biggest challenges that every metro-political areas is facing today. There are several tools and techniques evolved to predict pollution level which in turn helps in controlling and mitigating pollution. The three areas for computing pollution level are feature analysis, interpolation and prediction of fine grained air quality. These areas are providing extremely useful information so that one can take steps to mitigate pollution level, thus it also generates big societal impacts. Currently, there are individual models to address these issues separately. This paper proposes proposes single efficient framework by combining interpolation, prediction and feature analysis for air quality detection. This framework evaluates the different machine learning approaches to predict the air pollution components based on real data sets obtained from Bangalore. The main idea of the Urban Air Computing(UAC) is to gather the data of air quality and showing the feature analysis of air quality, interpolation and prediction