Multi-objective feature selection (MOFS) algorithms for prediction of liquefaction susceptibility of soil based on in situ test methods
| Autor: | Sarat Kumar Das, Ranajeet Mohanty, Mahasakti Mahamaya, Madhumita Mohanty |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
021110 strategic
defence & security studies Atmospheric Science 010504 meteorology & atmospheric sciences Artificial neural network Computer science 0211 other engineering and technologies Sorting Feature selection 02 engineering and technology 01 natural sciences Multi-objective optimization Cone penetration test Search algorithm Genetic algorithm Earth and Planetary Sciences (miscellaneous) Standard penetration test Algorithm 0105 earth and related environmental sciences Water Science and Technology |
| Zdroj: | Natural Hazards. 103:2371-2393 |
| ISSN: | 1573-0840 0921-030X |
| DOI: | 10.1007/s11069-020-04089-3 |
| Popis: | The prediction of liquefaction susceptibility for highly unbalanced database with limited and important input parameters is a crucial issue. The proposed multi-objective feature selection algorithms (MOFS) were applied to highly unbalanced databases of in situ tests: standard penetration test (SPT), cone penetration test (CPT) and shear wave velocity (Vs) test. Two multi-objective algorithms, non-dominated sorting genetic algorithm (NSGA-II) and multi-objective symbiotic organisms search algorithm (MOSOS), were coupled with learning algorithms, artificial neural network (ANN) and multivariate adaptive regression spline (MARS) separately to effectively select the optimal parameters and simultaneously minimize the error. The obtained optimal point has approximately equal accuracy in both liquefiable and non-liquefiable conditions for training and testing. The important inputs found for models based on SPT are: (N1)60, amax and Mw; CPT: qc1, amax and CSR and Vs: Vs1, CSR, amax and Mw. The CPT-based models were found to be the most efficient. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink