Identifying Earthquake Damages in Building Using Random Forest Classifier Algorithm

Detecting damaged buildings quickly after an earthquake is crucial for emergency teams to save lives. Existing methods for earthquake damage prediction in buildings face challenges related to accuracy and efficiency, particularly in quickly identifying damaged structures post-disaster. These challenges are overcome by the proposed approach using simple machine learning models to predict the stiff ness, strength, and displacement ratios of damaged materials based on observed crack features and loading conditions. This paper proposes an approach that combines classification by Decision Tree (DT) and regression tasks to calculate the damages using an earthquake prediction dataset. The numerical validation of the proposed method results in an accuracy of 0.893, a recall of 0.891, and an F1-score of 0.813, indicating its effectiveness. Compared to existing methods such as Artificial Neural Network (ANN), Support Vector Machine (SVM), VGG-16, and K-nearest neighbor (KNN), the proposed method overcomes limitations and enhances performance in identifying damaged structures.