Application of Genetic Algorithm Neural Network in Identifying Buildings in Landslide-Prone Areas

Indonesia is a disaster-prone country, one of which is landslides, which often occur in hilly areas with high rainfall. The impact damages the environment and infrastructure, especially buildings. For effective mitigation, a risk identification system based on artificial intelligence technology is needed. This study applies Genetic Algorithm Neural Network (GANN) in identifying buildings in landslide-prone areas. GANN was chosen for its ability to optimize network weights globally through selection, crossover, and mutation mechanisms, thus avoiding suboptimal local solutions. The dataset consists of 169 data with 12 structural features of the building. The model was configured with genetic parameters such as the number of generations 500, population size of 50, mutation rate of 10%, and the Stochastic Universal Sampling selection method. To Evaluate the performance of model created from dataset, we employed accuracy, precision, recall, and F1-score. The results showed an accuracy of 81% and an average F1-score of 0.82, with the best performance in the "Unsafe" class (recall 0.84). Although it still needs improvement, GANN has proven to have the potential as a decision support tool in data-driven landslide risk mitigation.