|
| Reference | Used method | Statistical parameter | Data division | Findings |
|
| [19] | ANN with training algorithms
ANN-LM
ANN-QN
ANN-GG
ANN-GD | R, RMSE, MAE, MAPE | 70% for training and 30% for testing.
106 data. | ANN-GG has the best prediction accuracy. |
| [20] | NN | R | 70% for training and 30% for testing.
233 data. | The results of the proposed model show great similarity with the test results |
| [3] | OSVM-AEW
LS-SVM with SOS | R, MAE, MAPE, RMSE | 67% for training and 33% for testing using a triple cross-validation approach. | OSVM-AEW has the best prediction accuracy |
| [21] | SVR-GA
SVR
ANN
GBDTs | RMSE, MAE, NSE, WI | 70% for training and 30% for testing.
217 test records. | SVR-GA gives better predictions than the other models. |
| [22] | GEP
ANN | R, RMSE, MAE | 70% for training, 10% for validation, and 20% for testing.
214 test records. | GEP has better prediction accuracy than ACI and CSA and shows very good agreement with the ANN model. |
| [4] | LS-SVR and SFA | R, RMSE, MAE, MAPE | 70% for training and 30% for testing.
214 data set | LS-SVR with SFA has better prediction accuracy compared to the standard SVR. |
| [23] | ANN | Mean, STD, COV | — | ANN performed better than the ACI code, EURO code, zsutty method, and russo method. |
| [24] | Strut-and-tie model | AVG, COV | 16 test specimens. | The results of the proposed model show great similarity with the test results. |
| [25] | CSTM | Mean, COV | 355 test specimens. | The performance of CSTM is better than other STM models. |
| [26] | New analytical expression using strut-and-tie model (STM) | Mean, COV | 111 test specimens. | The proposed model has better prediction accuracy compared to ACI 318-14 code. |
| [27] | GA | AVG, STD | 50% for training and 50% for testing.
371 data records. | GA has better prediction accuracy compared to ACI 318-14 code. |
| [28] | Feedforward NN | Mean, STD | 50% for training, 25% for validation, and 25% for testing.
433 data records. | The results of the proposed model are in agreement with the experimental and analytical data. |
| [29] | EMARS
BPNN
RBFNN
SVM | RMSE, MAPE | 10-fold cross-validation.
106 data set | The performance of EMARS is better than BPNN, RBFNN, and SVM. |
| [15] | RF
AdaBoost
GBRT
XGBoost | RMSE, MAE, MAPE | 10-fold cross-validation.
271 test records | The performance of the proposed models is better than the traditional machines single learning methods (DT, SVM, ANN) |
| [30] | SVR with RBF and polynomial kernel | R, RMSE | 10-fold cross-validation | SVR with RBF and polynomial kernel gives better prediction accuracy compared to backpropagation neural network and empirical relations. |
|
