Classifying Post-Eruption Housing Occupancy Status using Random Forest Algorithm

Abstract

Natural disasters such as the Mount Merapi eruption in Indonesia cause long-term damage to housing and settlement systems. After reconstruction, authorities must verify whether rebuilt houses are actually occupied to ensure fair aid distribution and effective recovery planning. This study applies the Random Forest algorithm to classify the occupancy status of post-eruption housing using real data collected by post-disaster rehabilitation agencies. We use a dataset of 2,516 housing records and split the data into 80% for training and 20% for testing. The experimental results show that the proposed model achieves an accuracy of 91.26% and an Area Under the Curve (AUC) value of 0.81. These results indicate that Random Forest performs well in distinguishing between occupied and unoccupied houses, even when the data reflect real-world imbalance and uncertainty. This study demonstrates that Random Forest is a practical and reliable machine learning method for post-disaster housing analysis. The model requires modest computational resources and handles mixed data types effectively. The findings confirm that ensemble-based machine learning can support government decision-making, strengthen monitoring of rehabilitation programs, and improve the overall effectiveness of disaster recovery efforts.