Building Damage Detection Based on Earthquake Impact and Gradient Boosting Method

Damage to buildings often occurs, and one of the causes is natural disaster such as earthquakes. Earthquakes frequently result in significant damage to buildings, causing financial losses due to damage to building facilities and even loss of life. Therefore, it is crucial to assess the damage to buildings to determine the extent of the damage. This research proposed an algorithm for detecting building damage using gradient boosting method. This method is similar to decision tree approach, but the decisions tree re-evaluated, resulting in smaller and more accurate data. For this analysis, the dataset was divided into two parts: training set and testing set. 80% of the dataset was used as training data, while 20% was used as testing data. After thorough data preprocessing, the gradient boosting method achieved an accuracy of 60.86% from large number of datasets compared to other methods, such as decision trees and random forests, the decision tree tends to overfit or underfit, especially with complex data. Meanwhile, the random forest method is generally faster and less prone to overfitting on large datasets. However, Gradient Boosting (GB) can achieve better accuracy, particularly for complex datasets. This result is indicating the effectiveness of the gradient boosting method.
Despite the large and complex dataset, where prediction results can sometimes vary, the outcomes demonstrate good performance. Future research should focus on refining datasets and optimizing the parameters used for predicting building damage.