Forecasting of outdoor air quality index using adaptive Neuro Fuzzy Inference System

Haripriyan Uthayakumar; Perarasu Thangavelu; Saravanathamizhan Ramanujam

doi:10.18502/japh.v6i3.8228

Haripriyan Uthayakumar Department of Chemical Engineering, A.C. Tech, Anna University, Chennai-600025, India
Perarasu Thangavelu Department of Chemical Engineering, A.C. Tech, Anna University, Chennai-600025, India
Saravanathamizhan Ramanujam Department of Chemical Engineering, A.C. Tech, Anna University, Chennai-600025, India

DOI: https://doi.org/10.18502/japh.v6i3.8228

Keywords: Adaptive Neuro-Fuzzy Inference System (ANFIS); Air pollution; Air Quality Index (AQI)

Abstract

Introduction: The estimation of air pollution level is well indicated by Air Quality Index (AQI), which tells how unhealthy the ambient air is and how polluted it can become in near future. Hence, the predictions or modeling of AQI is always of greater concern among researchers and this present study aims to develop such a model for forecasting the AQI.

Materials and methods: A combination of Artificial Neural Network (ANN) and Fuzzy logic (FL) system, called Adaptive Neuro-Fuzzy Inference System (ANFIS) have been considered for model development. Daily air quality data (PM2.5 and PM10) and meteorological data (temperature and humidity) over a period of March 2020 to March 2021 were used as the input data and AQI as the output variable for the ANFIS model. The performances of models were evaluated based on Root Mean Square Error (RMSE), Regression coefficient (R²) and Average Absolute Relative Deviation (AARD).

Results: A total of 100 datasets is split into training (70), testing (15) and simulation (15). Gaussian and Constant membership functions were employed for classifications and the final index consisted of 81 inference (IF/THEN) rules. The ANFIS Simulation result shows an R² and RMSE value of 0.9872 and 0.0287 respectively.

Conclusion: According to the results from this study, ANFIS based AQI is a comprehensive tool for classification of air quality and it is inclined to produce accurate results. Therefore, local authorities in air quality assessment and management schemes can apply these reliable and suitable results.