Neuro-fuzzy Model with Neighborhood Component Analysis for Air Quality Prediction

The issue of air contamination influencing wellbeing is a worldwide issue. The development of a health alarm system benefits not only the general public but also those requiring high levels of health surveillance. The challenge of this research is to develop a system that is highly accurate and improves its efficiency by selecting the appropriate features. This study presents the Neuro-fuzzy model by applying Neighborhood Component Analysis. The proposed model is used to classify the air quality index. The NCA method can specify past periods of data which high efficiency to optimize the prediction model performance. There are a total of 14 input features of meteorological and air pollution data from sensors plus short-term variation data. Various experiments are used to find an appropriate structure of the proposed model. In addition, four other different structures are utilized to confirm the efficacy of the proposed model. The result shows that the proposed model outperforms. Finally, the proposed model can be implemented to create a notification system to forecast air quality index up to three hours ahead.