Timeseries Forecasting of Delhi’s Air Quality Index Using Statistical and Neural Network Models
Abstract
Delhi consistently experiences severe air quality
episodes, with particulate loads frequently breaching recom-
mended limits. Anticipating the Air Quality Index (AQI) with
adequate lead time is central to timely advisories, exposure man-
agement, and responsive control actions. This work assembles a
comparative forecasting framework spanning statistical baselines,
machine learning, and deep sequence models to predict AQI
along with PM2.5 and PM10 (both in μg/m3
). The study evaluates
classical time-series tools (ARIMA, Prophet), non-linear regres-
sors (Support Vector Regression, Random Forest), and recurrent
neural networks (Long Short-Term Memory). Using five years of
hourly observations, we adopt a uniform pipeline for cleaning,
scaling, and strictly forward-in-time validation. Empirical results
show LSTM and Random Forest deliver consistent gains over the
statistical baselines, capturing rapid fluctuations and seasonal
shifts more faithfully. Overall, the analysis underscores the value
of hybrid, data-driven approaches for reliable urban air quality
forecasting and supports targeted mitigation in highly polluted
settings.
References
H. Kumar, M. P. Singh, and R. Gupta, “A Hybrid CNN LSTM Model for Spatiotemporal PM2.5 Concentration Forecasting in Delhi,” IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1–12, 2022.
A. Sharma, S. Goyal, and A. Singh, “Machine Learning for Air Quality Forecasting: A Comprehensive Review,” Journal of Cleaner Production, vol. 358, p. 132035, 2022.
R. Verma and P. Nagpal, “An Attention-Based Spatiotemporal Graph Convolutional Network for Air Quality Prediction in Urban India,” Atmospheric Pollution Research, vol. 14, no. 5, p. 101734, 2023.
S. Jain, A. Patel, and S. Kumar, “Impact of Meteorological Parameters on Air Quality Index Prediction Using Deep Learning,” in Proc. IEEE Int. Conf. on Big Data, Sorrento, Italy, 2023, pp. 2451–2460.
P. Li, Y. Wang, and Z. Zhang, “Transformer-Based Models for Time Series Forecasting: A Survey and Taxonomy,” ACM Computing Surveys, vol. 55, no. 11, pp. 1–37, 2023.
D. Gupta and S. Singh, “Forecasting PM10 Concentrations in India: A Comparative Analysis of ARIMA, Prophet, and XGBoost,” Environmen- tal Science and Pollution Research, vol. 29, pp. 78910–78925, 2022.
Central Pollution Control Board (CPCB), India. [Online]. Available: https://cpcb.nic.in/
“Air Quality Data in India (2015–2020),” Kaggle. [Online]. Available: https://www.kaggle.com/datasets/rohanrao/air quality-data-in-india
Y. Zhang, J. Chen, and L. Liu, “Hybrid LSTM–CNN Model for Air Quality Prediction Integrating Meteorological Parameters,” Atmospheric Pollution Research, vol. 14, no. 5, p. 101255, 2023.
J. Li and P. Wang, “Meteorology-Driven Deep Learning for PM2.5 Forecasting: A Hybrid Approach,” Environmental Modelling & Software, vol. 157, p. 105473, 2022.
