|
Air quality index of New Delhi forecasting |
|---|---|
| รหัสดีโอไอ | |
| Title | Air quality index of New Delhi forecasting |
| Creator | Chidchanok Ratakai |
| Contributor | Jirachai Buddhakulsomsiri, Advisor |
| Publisher | Thammasat University |
| Publication Year | 2568 |
| Keyword | Forecast, Regression, Machine learning, Neural network, Air quality index (AQI) |
| Abstract | Air pollution remains a critical environmental issue in many urban areas, with New Delhi frequently experiencing AQI levels that pose risks to public health and daily activities. Reliable forecasting of air quality conditions can assist authorities and communities in responding to pollution episodes before they become severe. This study develops forecasting models for weekly Air Quality Index (AQI) prediction in New Delhi using statistical, machine learning, and ensemble forecasting approaches. Daily AQI observations covering the period from April 2022 to April 2025 were collected and transformed into weekly averages to reduce short-term variability and emphasize broader temporal patterns. The resulting dataset consisted of 161 weekly observations and was divided into training, validation, and testing subsets. Four forecasting approaches were investigated, namely Seasonal Autoregressive Integrated Moving Average (SARIMA), Holt’s Winter exponential smoothing, Artificial Neural Networks (ANN), and ensemble forecasting models. The forecasting results indicated that SARIMA and Holt’s Winter models could represent the seasonal characteristics present in the AQI series, while the ANN model provided an alternative approach for capturing nonlinear relationships within the data. To further improve forecasting performance, forecasts from the individual models were combined using ensemble techniques. Among the approaches evaluated, the weighted ensemble model achieved the strongest predictive performance and generated the lowest forecasting errors. The findings suggest that integrating forecasts obtained from different analytical approaches can enhance prediction accuracy and improve forecasting reliability. The proposed framework provides useful support for environmental monitoring, public health planning, and air quality management. In addition, the study demonstrates the potential benefits of ensemble forecasting for AQI prediction and offers a foundation for future research involving advanced forecasting techniques and larger environmental datasets. |
