Wrapper subset evaluation facilitates the automated detection of the ground-motion intensity measures and derivation of the seismic fragility curves

Fragility analysis and its graphical representation in terms of fragility curve is an effective tool for seismic risk assessment of structural systems. Fragility curve is a statistical tool representing the probability of exceeding a given damage state as a function of a ground-motion intensity measure (IM) that represents the ground motion. One of the most important issues in deriving fragility curves is the large number of IMs that have been proposed by researchers over the years. Any improvement in the proper selection of the IM would therefore represent a significant gain with respect to accurately predicting the structural seismic responses and mitigating the economic impacts of earthquake disasters besides
Reducing the loss of lives. In this study we apply automated machine learning to analyse IMs and assessing the seismic responses of two typical one- and two-storey unreinforced masonry (URM) buildings located in the Kurdistan region of Iraq. By applying wrapper feature selection method and using several classifier algorithms it was able to select the most relevant IMs to use as input to a fragility analysis tool. Furthermore, results suggest that the prediction of the failure pattern of buildings is also feasible using the wrapper method