Estimation of fundamental period of concrete frames with infill walls using decision tree

Fundamental period of vibration of a structure is an important parameter which goes in the seismic analysis and design. There are limitations in applying analytical or semi-empirical formulations provided in various design codes and research articles, arising due to the geometric and material variations, distribution of seismic mass in plan and elevation, infill walls, and opening in walls. Soft computing or machine learning techniques could be applied to achieve better prediction accuracy of the estimated fundamental period. Artificial neural network (ANN) models have been reported to be successful in predicting the behaviour of such infilled frames. Tree-based approach using decision tree (DT) algorithm is explored in this article for the same purpose, and with excellent results. The performances of DT models were independent of the data divisions, or the percentage of openings in the brick infill walls. The accuracy of the DT models was exemplified to be superior to the empirical expressions from the codes/standards (an updated summary is provided in the article) across the world. Moreover, the accuracy of the estimates obtained with DT models was comparable to the other soft computing models (ANN) reported in the literature.