Enhancing Rainfall Prediction Accuracy through XGBoost Model with Data Balancing Techniques

Data sampling in machine learning pipelines significantly impacts model performance. This research explores the repercussions of inaccurate sampling on model accuracy, revealing substantial degradation, especially for rare or underrepresented data points. Various datasets and sampling methods were employed, highlighting the critical role of meticulous sampling in constructing reliable, generalizable models. In a study focused on IoT weather classification, five data balancing methods were tested, showing that Random Oversampling effectively improved prediction accuracy, achieving a metric score performance of 99%. The results emphasize the importance of addressing class imbalance for robust machine learning predictions. Specifically, XGBoost demonstrated exceptional performance with a 100% accuracy rate in an IoT-based weather detection algorithm, showcasing its effectiveness in predicting weather patterns. These findings provide valuable insights for researchers and practitioners across domains relying on data-driven models for accurate predictions and informed decision-making.