Multiple artificial intelligence (AI) algorithms were tested using the database to determine the optimal model for predicting the DI. The ANN LM 4-21-1 model emerged as the best-fit model, achieving high accuracy with an a10-index of 0.9888 for training data and 0.9911 for test data. The R index, which measures the goodness of fit, was found to be 0.9996 and 0.9995, respectively, indicating a strong correlation. The smoothness of the fitted curves suggests that overfitting was avoided.

The study emphasized the significance of considering both the wall width and masonry tensile strength in the analysis, as neglecting these factors could lead to unreliable results in risk assessments. Based on the selected AI model, an equation was derived to calculate the DI, enabling its direct integration into risk assessment equations. However, it is important to acknowledge the limitations and constraints of the results, such as database limitations and challenges in characterizing the structural typology, as discussed in the paper. Further validation of the results through comparison with real-world cases is recommended, which can be achieved by applying the proposed equation.