Wavelet-Based Combinatorial Machine Learning Techniques for Forecasting Particulate Matter Concentrations

Abstract

Air pollution in Delhi is so severe that, in most years, the city's average daily air quality index (AQI) consistently falls below the “Poor” category. Efficient and reliable prediction models can provide a guide to effective air pollution control. In this study, daily concentrations of PM2.5 and PM10 (particle sizes ≱ 2.5 μm and ≱ 10 μm, respectively) in a residential area (Alipur), an industrial area (Okhla), and a traffic area (Pusa) of Delhi have been modeled using wavelet-based machine learning (ML) algorithms. Wavelet methodology with the help of filter bank decomposes a series into various sub-parts to gather information at multiresolution levels. Performance of various important wavelet filters namely, haar, Daubechies (d4), Coiflet (c6), best-localized Daubechies (bl14), and least asymmetric (la8) have been recorded to obtain the best combination for forecasting particulate matter (PM) levels one week and one month in advance. The proposed models have performed well in predicting PM concentration.

Received: 26 April 2025 | Revised: 2 September 2025 | Accepted: 22 October 2025

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

Data are available from the corresponding author upon reasonable request.

Author Contribution Statement

Sandip Garai: Conceptualization, Methodology, Software, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization. Debopam Rakshit: Conceptualization, Methodology, Software, Formal analysis, Investigation, Resources, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration. Arkaprava Roy: Conceptualization, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization. Ranjit Kumar Paul: Conceptualization, Methodology, Validation, Investigation, Resources, Writing – original draft, Writing – review & editing, Supervision, Project administration. Ritwika Das: Conceptualization, Validation, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization.