GAN-Based Deep Learning Framework for Climate Forecasting in Smart Cities

Temperature is nature’s pulse, influencing our decisions, protecting our health, and shaping the world. Temperature and cloud cover are inversely correlated, which means that increased cloud cover typically leads to lower temperatures, and decreased cloud cover allows for higher temperatures. Accurate temperature and cloud cover predictions are crucial as these factors significantly influence daily life. This study utilises historical data to predict temperature and cloud cover for Hyderabad, a smart city. Generative Adversarial Networks (GANs) are utilised for data imputation to handle missing values effectively. Subsequently, Gated Recurrent Unit (GRU) and Bidirectional GRU (BiGRU) models are employed for predictive analysis. The results demonstrate that the BiGRU model outperforms the GRU model, achieving a normalized mean absolute error (nMAE) of 0.15 before data imputation and 0.12 after imputation for a single parameter, indicating improved prediction accuracy.