A Hybrid Ensemble Prediction Method for Analyzing Air Quality Data

Air pollution is one of the biggest environmental challenges that world is facing in this era. The proposed work aims to address the problem of urban air pollution by analyzing the trends of urban air pollution data effectively. The past air pollution trends are captured in terms of time series data of pollutants. Hence, a partitioning-based hybrid approach is utilized to identify similar time series consisting of nitrogen dioxide concentration for various locations of Delhi. In addition to forecasting the time series data using already existing methods, exponential smoothing is further performed to capture the overall historical effect of the time series. Further, using this method, time series forecasting is performed for the further intervals of nitrogen dioxide concentrations. Evaluation of our approach is done on various metrics such as RMSE and MAPE, and optimistic results are obtained.