Investigation into the robustness of evolutionary programming regression for sedimentation study
Evolutionary Polynomial Regression (EPR) has been used to determine the total sediment load in selected rivers in Malaysia. In order to test the robustness and generalization ability of EPR modelling, the approach that is generally adopted is to test the performance of trained EPR mod...
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Asian Research Publishing Network (ARPN)
2016
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/21731/ http://umpir.ump.edu.my/id/eprint/21731/ http://umpir.ump.edu.my/id/eprint/21731/1/Investigation%20into%20the%20robustness%20of%20evolutionary%20programming.pdf |
| Summary: | Evolutionary Polynomial Regression (EPR) has been used to determine the total sediment load in selected rivers
in Malaysia. In order to test the robustness and generalization
ability of EPR modelling, the approach that is generally
adopted is to test the performance of trained EPR models on an independent validation set. If such performance is adequate, the model is deemed to be robust and able to generalize. When evaluating EPR models, consideration must be given not only to their predictive accuracy but also to the interpretive ability of the models. This can be done by carrying out a sensitivity analysis that quantifies the relative importance of model inputs to the corresponding outputs. In this paper, the robustness of EPR models is investigated in a case study of predicting the total sediment load at Malaysian rivers. A procedure that tests the robustness of the predictive ability of EPR models is introduced. The results indicate that the good performance of EPR models in the data used for model calibration and validation also perform in a robust fashion over a range of data used in the model calibration phase. The results also indicate that validating EPR models using the procedure applied in this study are essential in order to investigate their robustness. |
|---|