Forecasting PM10 concentration in Iraq based on improving v-support vector regression

Abstract

This manuscript develops and evaluates an improved v-support vector regression (v-SVR) framework, optimized via the Bamboo Forest Growth Optimization (BFGO) algorithm, to forecast daily PM10 concentrations in Baghdad, Iraq. Daily PM10 data from a ground-based monitoring station at the Baghdad U.S. Embassy, covering the period 1 March 2019 to 1 June 2023, are combined with relevant meteorological and environmental predictors to construct a data-driven forecasting system. The proposed method simultaneously performs v-SVR hyperparameter tuning and feature selection, addressing limitations of conventional v-SVR schemes and existing nature-inspired approaches that typically optimize only model parameters. BFGO is tailored to guide the search in the v-SVR hypothesis space through meme-based grouping, whip extension (exploitation), and shoot growth (exploration), thereby enhancing convergence toward a parsimonious and high-performing regression model. Model performance is assessed under multiple forecast horizons using standard accuracy indices such as RMSE, MAE, and R${}^{2}$, and is benchmarked against baseline v-SVR, Random Forest, and other machine-learning forecasters commonly used in air quality prediction. The empirical results show that BFGO-v-SVR consistently achieves lower prediction errors and higher explanatory power than competing models, particularly in capturing sharp peaks and rapid fluctuations in PM10 driven by dust storms and intense anthropogenic emissions. These gains are attributed to the joint optimization of kernel parameters, regularization strength, $\epsilon$-insensitive loss width, and an informative subset of predictors, which together improve generalization and reduce overfitting. The proposed BFGO-v-SVR framework offers a robust and flexible tool for real-time PM10 forecasting in highly polluted urban environments and can support more effective air quality management, early-warning systems, and evidence-based environmental policy in Iraq and similar semi-arid regions.