| Reference | Number of data samples | Method and input | Performance criteria for testing part | R2 | RMSE (kN) | MAE (kN) |
| Amayreh and Saka [36] | 47 | 8 inputs (minimum web yield stress, span of the castellated beam, overall depth, minimum width of the web post, web thickness, flange thickness, width of flange, loading condition) Method: backpropagation network | 0.995 | Not provided | Not provided |
| Gholizadeh et al. [37] | 140 | Geometrical inputs only (the overall depth of castellated beam; the vertical projection of inclined side of opening; the web thickness; the flange width; the flange thickness; the width of web post at middepth; the horizontal projection of inclined side of opening) | | | | BP1 | | 4.0625 | | BP2 | | 3.5611 | | ANFIS1 | | 2.7276 | | ANFIS2 | | 2.0631 | |
| Gandomi et al. [38] | 47 | 8 inputs (minimum web yield stress, span of the castellated beam, overall depth, minimum width of the web post, web thickness, flange thickness, width of flange, loading condition) | | | | Methods | | | | Genetic programming | 0.817 | | 32.33 | Least-squares regression (LSR) | 0.682 | | 36.60 |
| Aminian et al. [39] | 142 | 5 inputs (the overall depth of castellated beam, the vertical projection of inclined side of opening, the web thickness, the minimum web yield stress, and the width of web post at middepth) | | | | Linear genetic programming | 0.960 | | 4.62 | GSA algorithm | 0.952 | | 4.95 |
| This investigation | 150 | 9 inputs (the overall depth of castellated beam, the vertical projection of inclined side of opening, the web thickness, the flange width, the flange thickness, the width of web post at middepth, the horizontal projection of inclined side of opening, the minimum web yield stress, the minimum flange yield stress) | 0.989 | 3.328 | 2.622 |
|
|