Share
Export Citation
An Integration of Tomek Links as Cleaning Step and SMOTE for Handling Imbalanced Class in Rainfall Classification
Abidin N.A.S.Z.
Proceedings of SPIE the International Society for Optical Engineering
Abstract
Rainfall can affect human activities, therefore very high rainfall intensity can cause various problems, such as natural disasters that can damage ecosystems and problems in various fields. Therefore, rainfall is an important thing to study further, especially in terms of the accuracy and performance of its predictions. Machine learning can help identify rainfall patterns based on historical data. However, in practice, there is a problem, namely class imbalance, often referred to as imbalanced class. Imbalanced class is a situation where the distribution of existing data is uneven. An imbalanced class can affect the accuracy of the model, resulting in classification errors. This situation also causes machine learning algorithms to be biased towards the majority class. This study introduces Tomek links, which play a role in the cleaning step, and SMOTE as a method to overcome imbalanced class. Based on research, this method produces an accuracy of 92.96% using Tomek links as a cleaning step and SMOTE, and produces an accuracy of 84.84% without using these methods. This shows that the proposed method successfully addresses class imbalance and improves model accuracy in rainfall classification cases.