Utilizing Support Vector Machine (SVM) to predict relative humidity in selected Ugandan cities

Abstract

This study focuses on implementing a Support Vector Machine (SVM) model to predict relative humidity (RH) in selected cities of Uganda, addressing the critical gap in tailored predictive models for Uganda's unique climatic conditions. Leveraging three years of monthly RH data from the Uganda National Meteorological Authority, the SVM model demonstrates significant nonlinearity in RH data. Results reveal varying performance across different towns and times, with higher accuracy observed in training data (58%) compared to testing data (33%). Kampala and Arua exhibit the highest RH, with better prediction performance. The study underscores the potential of SVM for RH prediction in Uganda, offering valuable insights for sectors such as agriculture, health, and urban planning. Challenges in generalizing model performance to unseen data are noted, suggesting avenues for future research to enhance model robustness and applicability in the Uganda's context.