Journal of Building Engineering 63 (2023) 105493 Available online 4 November 2022 2352-7102/© 2022 Elsevier Ltd. All rights reserved. An interpretable machine learning method for the prediction of R/ C buildings’ seismic response Konstantinos Demertzis a, b , Konstantinos Kostinakis c, * , Konstantinos Morfdis d , Lazaros Iliadis b a School of Science & Technology, Informatics Studies, Hellenic Open University, Greece b School of Engineering, Department of Civil Engineering, Faculty of Mathematics Programming and General Courses, Democritus University of Thrace, Kimmeria, Xanthi, Greece c Department of Civil Engineering, Aristotle University of Thessaloniki, Aristotle University Campus, 54124, Thessaloniki, Greece d Earthquake Planning and Protection Organization (EPPO-ITSAK), Terma Dasylliou, 55535, Thessaloniki, Greece A R T I C L E INFO Keywords: Interpretable machine learning Model validation Seismic damage prediction Structural vulnerability assessment Reinforced concrete buildings ABSTRACT Building seismic assessment is at the forefront of modern scientifc research. Several researchers have proposed methods for estimating the damage response of buildings subjected to earthquake motions without conducting time-consuming analyses. The advancement of computer power has resulted in the development of modern soft computing methods based on the use of Machine Learning (ML) algorithms. However, a lack of expertise associated with the use of complex ML architectures can affect the performance of the intelligent model and, ultimately, reduce the al- gorithm’s reliability and generalization which should characterize these systems. The current paper proposes a fully validated interpretable ML method for predicting seismic damage of R/C buildings. Specifcally, the most effcient machine learning algorithms were used in a large-scale comparison study in a sophisticated dataset of 3D R/C buildings. Moreover, effective additional validation ensures that models are sound, have low complexity, are fair and provide clear ex- planations for decisions made. Also, extensive experiments were done to make the fnal machine learning model explainable and the decisions interpretable. The proposed method aims to suggest that the civil protection mechanisms must include scientifc methodology and appropriate tech- nical tools into their technological systems, in order to make substantial innovative leaps in the new era. 1. Introduction One of the most important, but also challenging, scientifc issues in the feld of earthquake engineering is the estimation of the structural response of buildings subjected to earthquake ground motions. Since now, numerous research studies have dealt with the above issue and proposed a vast variety of different methods aiming at the seismic assessment of structures. Many of these methods focus on the rapid determination of the earthquake damage response and on the seismic vulnerability assessment of large number of buildings without performing computationally hard analyses, in an attempt to overcome the diffculties resulting from the time- consuming conduction of demanding nonlinear analysis methods (e.g Refs. [1–6]), These procedures usually utilize methods based on the application of statistics theory. In the last decades, the increase of the computers’ power has led to the development of modern * Corresponding author. E-mail address: kkostina@civil.auth.gr (K. Kostinakis). Contents lists available at ScienceDirect Journal of Building Engineering journal homepage: www.elsevier.com/locate/jobe https://doi.org/10.1016/j.jobe.2022.105493 Received 21 July 2022; Received in revised form 13 October 2022; Accepted 29 October 2022