<![CDATA[Abstract−Landslides are one of the most significant natural disasters in Indonesia, often causing substantial economic losses and threats to human safety. A key challenge in processing landslide data is the issue of class imbalance, where the number of disaster occurrence data is significantly smaller compared to non-disaster data. This study aims to improve landslide prediction accuracy by integrating the Naive Bayes algorithm and Particle Swarm Optimization (PSO) while employing the Random Undersampling (RU) technique to address data imbalance. The dataset used in this study includes landslide data from Samarinda City for the period 2022-2023, obtained from the Regional Disaster Management Agency (BPBD) and the Meteorology, Climatology, and Geophysics Agency (BMKG). The research process involved data preprocessing, balancing data using RU, implementing the Naive Bayes algorithm, and optimizing it with PSO. Model performance was evaluated using the 10-Fold Cross Validation technique and a confusion matrix. The results show that applying the Naive Bayes algorithm with PSO optimization without RU achieved the highest average accuracy of 89.49%, compared to Naive Bayes without optimization, which only reached 87.59%. Meanwhile, the application of RU showed varied effects, with the combination of Naive Bayes + PSO with RU achieving an average accuracy of 50%, slightly better than Naive Bayes with RU, which only reached 45%. This study demonstrates that PSO optimization can improve the performance of the Naive Bayes model in handling complex landslide datasets, although balancing techniques such as RU must be applied cautiously to avoid the loss of important information. The results of this study are expected to support disaster mitigation efforts through more accurate predictions, aiding stakeholders in decision-making, such as early evacuation planning and infrastructure development in landslide-prone areas.]]>
