Optimasi Model Random Forest untuk Memprediksi Coral Bleaching dengan Penerapan Komputasi Paralel

Untuk Memprediksi Coral Bleaching

Abstract

Coral bleaching is a critical environmental issue caused by environmental stressors, such as rising sea temperatures, which result in the loss of algae symbiosis within corals. However, predicting coral bleaching remains challenging due to the complexity of environmental conditions, the uncertainty of contributing factors, and the limited availability of accurate and consistent data. Additionally, managing large datasets and ensuring efficient training of predictive models with complex datasets pose significant challenges. This study explores the application of parallel computing in developing a predictive Random Forest model to forecast coral bleaching events based on environmental data, including sea surface temperature (SST), sea surface temperature anomalies (SSTA), depth, and location coordinates. Parallel computing is employed to enhance efficiency in training the model by utilizing multi-core processors, significantly reducing execution time. The results demonstrate that the model achieves a prediction accuracy of 95.19% with an R-squared value of 0.685. The application of parallel computing also shows a reduction in computation time, although not always linear due to the overhead associated with task management. This research is expected to support coral reef conservation efforts by providing a faster and more accurate predictive model.

Keywords – Parallel Computing; Random Forest; Coral Bleaching.