<![CDATA[The demand for fast, efficient, and adaptive emergency housing continues to increase, especially in disaster-prone areas and large-scale displacement situations. The determination of the design of Temporary Modular Shelter (TMS) so far still depends a lot on subjective considerations, so a more systematic and data-based approach is needed. This study develops and validates an Artificial Neural Network (ANN) model to identify the most suitable TMS design based on performance indicators and expert assessment. The approach was carried out through the Systematic Literature Review (SLR) stage, the determination of eight key design indicators, and assessment by 150 multidisciplinary respondents. The ANN model was built using a dense four-layer architecture with a total of 1,780 parameters and trained for 400 epochs using the TensorFlow and Keras libraries. The results showed a validation accuracy of 96% and a macro F1-score of 0,9146, indicating the stability and reliability of the model. Analysis of the contribution of features with the SHAP method revealed that the indicators of assembly methods, availability of human resources, and availability of local materials had the greatest influence on the classification results. This model has proven to be effective as a decision support system that is able to increase objectivity and efficiency in the TMS design process. Further development is suggested through integration into web-based digital platforms or mobile applications to support rapid and adaptive emergency response planning.]]>
