sc 21(14): e4

Research Article

Spatio-temporal Prediction of Air Quality using Distance Based Interpolation and Deep Learning Techniques

Download147 downloads
  • @ARTICLE{10.4108/eai.15-1-2021.168139,
        author={K. Krishna Rani Samal and Korra Sathya Babu and Santos Kumar Das},
        title={Spatio-temporal Prediction of Air Quality using Distance Based Interpolation and Deep Learning Techniques},
        journal={EAI Endorsed Transactions on Smart Cities},
        volume={5},
        number={14},
        publisher={EAI},
        journal_a={SC},
        year={2021},
        month={1},
        keywords={Air quality, Deep Learning, LSTM, Inverse Distance Weighting, Spatio-temporal prediction},
        doi={10.4108/eai.15-1-2021.168139}
    }
    
  • K. Krishna Rani Samal
    Korra Sathya Babu
    Santos Kumar Das
    Year: 2021
    Spatio-temporal Prediction of Air Quality using Distance Based Interpolation and Deep Learning Techniques
    SC
    EAI
    DOI: 10.4108/eai.15-1-2021.168139
K. Krishna Rani Samal1,*, Korra Sathya Babu1, Santos Kumar Das1
  • 1: National Institute of Technology, Rourkela, India
*Contact email: 517cs6019@nitrklac.in

Abstract

The harmful impact of air pollution has drawn raising concerns from ordinary citizens, researchers, policy makers, and smart city users. It is of great importance to identify air pollution levels at the spatial resolution on time so that its negative impact on human health and environment can be minimized. This paper proposed the CNN-BILSTM-IDW model, which aims to predict and spatially analyze the pollutant levelin the study area in advance using past observations. The neural network-based Convolutional Bidirectional Long short-term memory (CNN-BILSTM) network is employed to perform time series prediction over the next four weeks. Inverse Distance Weighting (IDW) is utilized to perform spatial prediction. The proposed CNN-BILSTM-IDW model provides almost 16% better prediction performance than the ordinary IDW method, which fails to predict spatial prediction at a high temporal period. The results of the presented comparative analysis signify the efficiency of the proposed model.