Closure to “An integrated framework of extreme learning machines for predicting scour at pile groups in clear water condition” by: I. Ebtehaj, H. Bonakdari, F. Moradi, B. Gharabaghi, Z. Sheikh Khozani

Abstract The authors of the paper under discussion predicted the scour depth around pile groups in clear water using the extreme learning machine (ELM) neural network. The ELM model performance results were compared with those obtained with support vector machine (SVM) and artificial neural network (ANN). This discussion highlights Baghbadorani, Ataie-Ashtiani, Beheshti's (2018) incorrect calculations in Table 3 and scatter plots in Fig. 1 presented in their discussion. For example, they indicated the overall performance of R2 is −2.5 and MAPE is 67% for the 27 raw data listed in the table. They reported these statistical indices not only according to erroneous calculations but also for merely 27 samples selected arbitrarily and without valid justification from over 1 100 data available from previous studies. Using 27 out of 1 100 data points is evidently not a sound means of evaluating the overall accuracy of the proposed equation. Subsequently, the authors provide information to further clarify their responses to the comments in the discussion paper.