Research on Multimodal Air Quality Index Prediction Method Based on Hybrid VLOA-NM Optimized VMD and IWOA-LSTM

Predicting the Air Quality Index (AQI) carries irreplaceable scientific significance and practical value for safeguarding public health, guiding individual protective behaviors, and optimizing regional pollution control strategies. To further enhance the prediction accuracy for AQI, a novel hybrid model integrating the Vampire Leech Optimization Algorithm (VLOA), Nelder-Mead local search, Variational Mode Decomposition (VMD), Whale Optimization Algorithm (WOA), and Long Short-Term Memory (LSTM) network was proposed. Initially, the combination of VLOA with Nelder-Mead local search was employed to optimize the VMD algorithm, effectively reducing the non-stationarity of AQI data. Furthermore, the LSTM neural network demonstrates exceptional performance in addressing the nonlinear coupling between meteorological conditions lag effects and pollution accumulation processes. After optimizing the LSTM with the improved Whale Optimization Algorithm, the model exhibited enhanced capability to capture long-range dependencies, significantly improving prediction accuracy. Experimental results indicated that the proposed VNM-VMD-IWOA-LSTM model achieved lower prediction errors compared to other benchmark models, with a coefficient of determination ($R^{2}$) reaching 95.43%. This finding demonstrates that the VNM-VMD-IWOA-LSTM model can effectively predict AQI variations, offering a promising alternative methodology for air quality forecasting in support of public health initiatives.