Knowledge extraction from trained neural network scour model at culvert outlets
Artificial neural networks (ANNs), due to their outstanding capabilities for modeling complex processes, have been successfully applied to a variety of problems in hydraulics. However, one of the major criticisms of ANNs is that they are just black-box models, since a satisfactory explanation of their behavior has not been offered. They, in particular, do not explain easily how the inputs are related to the output and also whether the selected inputs have any significant relationship with an output. In this paper, a perturbation analysis for determining the order of influence of the elements in the input vector on the output vector is discussed. The analyses of the results suggest that each variable in the input vector (d 50/d 0, F 0, H/d 0, σg, and W 0/d 0) influences the depth of scour in different ways. However, the magnitude of the influence cannot be clearly quantified by this approach. Further it adds that the selection of input vector based on linear measures between the variables of interest, which is commonly employed, may still include certain spurious elements that only increase the model complexity.
KeywordsArtificial neural networks Scour depth Culvert outlets
The authors gratefully acknowledge the financial support from the International Foundation for Science (IFS) and Organisation for the Prohibition of Chemical Weapons (OPCW) as research grant (Grant no. W/5073-1). Authors also express their sincere gratitude to Universiti Sains Malaysia (USM) and Ministry of Higher Education (MOHE), Malaysia for allowing necessary financial support through research grants (USM Short Term Grant, Code no P3665; and Explanatory Research Grant Scheme-ERGS, code no. X0043).
- 2.Azamathulla HMd, Haque MA (2012) Prediction of scour depth at culvert outlets using gene-expression programming. Int J Innov Comput Inform Control 8(7B):5045–5054Google Scholar
- 4.Azamathulla HMd, Ghani AAb, Zakaria NA, Chang CK, Leow CS (2008) Knowledge extraction from trained neural network scour models. Mod Appl Sci Can 2(4):52–62Google Scholar
- 8.Eberhart RC, Dobbins RW (1990) Neural network PC tools: a practical guide. Academic, New YorkGoogle Scholar
- 9.Gandomi AH, Alavi AH (2011) Applications of computational intelligence in behavior simulation of concrete materials. In: Yang XS, Koziel S (eds) Chapter 9 in Computational optimization and applications in engineering and industry. Springer, New York SCI, vol 359, pp 221–243Google Scholar
- 10.Gandomi AH, Yang XS, Talatahari S, Alavi AH (2013) Metaheuristic applications in structures and infrastructures. Elsevier, AmsterdamGoogle Scholar
- 11.Ho YC (1992) Perturbation analysis: concepts and algorithms.In: Arlington VA, Swain JJ, Goldsman D, Crain RC, Wilson JR (eds) Proceedings of the 1992 winter simulation conference, pp 231–240Google Scholar
- 16.Lozowski A, Cholewo TJ, Zurada JM (1996) Crisp rule extraction from perception network classifiers. In: Proceedings of the IEEE international conference on neural networks. Plenary, Panel, and Special Sessions, Washington, DC, pp 94–99Google Scholar
- 17.Maier HR, Dandy GC (2000) Neural networks for the prediction and forecasting of water resources variables: a review of modelling issues and applications. Environ Model Softw 15(1):101–124Google Scholar
- 25.Yang XS, Gandomi AH, Talatahari S, Alavi AH (2012) Metaheuristis in water, geotechnical and transportation engineering. Elsevier, AmsterdamGoogle Scholar