Data for: An Automated Machine-Learning-Assisted Stochastic-Fuzzy Multi-Criteria Decision Making Tool: Addressing Record-to-Record Variability in Seismic Design

The multi-criteria decision making (MCDM) dataset is presented for robust-to-uncertainty seismic dam design. It is associated with the “Appendix D Earthquake Engineering MCDM Dataset: Robust-to-Uncertainty Seismic Design” of a paper published in the Applied Soft Computing journal titled “An Automated Machine-Learning-Assisted Stochastic-Fuzzy Multi-Criteria Decision Making Tool: Addressing Record-to-Record Variability in Seismic Design”. Here, the results of the AutoML-assisted net outranking flow predictions (Φnet) for the design of experiments (DoEs) of [2000, 1000], considering all 20 alternatives and 18 ML techniques during intelligent dam design, are accessible. As can be seen in the “Earthquake-Engineering-MCDM-dataset” Excel file, there are three spreadsheets: (i) Fuzzy Numbers; (ii) DoE=1000; and (iii) DoE=2000. In the “Fuzzy Numbers” spreadsheet, columns A and B denote alternative descriptions, and columns C to DB represent triangular trapezoidal fuzzy numbers (TFNs) for the calculated criteria. TFNs for the structural performance, design safety, and construction cost criteria are shown in columns C-AX, AY-CT, and CU-CX, respectively. For example, columns C-F represent TFNs for the first criterion (C1), G-J for the second one (C2), and CU-CX for the last criterion (C25). The first 125 columns in the “DoE=1000” spreadsheet, columns A-DU, represent input parameters for the developed MCDM tool, while the last 20 columns, DV to EO, refer to the outputs. In detail, the first, second, third, fourth, and fifth groups of 25 columns are the generated criteria weights, preference functions, and preference thresholds (p, q, and s), respectively. The 20 alternatives, indicated by Φnet as our desired outputs, are shown in columns DV-EO. A similar description can be repeated for the “DoE=2000” spreadsheet, in which the input-output number of rows increased from 1000 to 2000, respectively, in this scenario. Regarding the AutoML-aided predictions, the calculated error metrics are introduced in the “ML-Metrics-DoE1000.xlsx” and “ML-Metrics-DoE2000.xlsx” Excel files for DoEs of [1000, 2000], respectively. Each of these files includes 21 spreadsheets, in which the first 20 ones report the AutoML step performance metrics for each alternative, while the last one summarizes all the results. In the first 20 spreadsheets, columns A and B denote ML algorithms, whereas columns C-I stand for the calculated MAE, MSE, RMSE, R2, RMSLE, MAPE, and training time (TT), respectivel