Influence of neural network structure and data-set size on its performance in the prediction of height of growth hormone-treated patients

• Autorzy: Smyczyńska U. , Smyczyńska J. , Tadeusiewicz R.

Identyfikatory

DOI

10.1515/bams-2016-0001

• Języki publikacji: EN

Abstrakty

EN

It is well known that the structure of neural network and the amount of available training data influence the accuracy of developed models; however, the exact character of this relation depends on the chosen problem. Thus, it was decided to analyze what impact these parameters have on the solution of the problem on which we work – the prediction of final height of children treated with growth hormone. It was observed that multilayer perceptron with a wide range of numbers of hidden neurons (from 1 to 100) could solve the problem almost equally well. Thus, this task seems to be rather simple, not requiring complex models. Larger networks tended to produce less accurate results and did not generalize well while working with the data not used in training. Repeating the experiment with the training data set reduced to 50% of its original content, as expected, caused a decrease in accuracy.

Słowa kluczowe

EN

impact; data-set size; MLP model; multilayer perceptron; optimization; neural network structure; treatment efficiency prediction

• Wydawca: Uniwersytet Jagielloński - Collegium Medicum ; Index Copernicus Sp. z o.o.
• Czasopismo: Bio-Algorithms and Med-Systems
• Rocznik: 2016
• Tom: Vol. 12, no. 2
• Strony: 53--59
• Opis fizyczny: Bibliogr. 18 poz., rys., wykr.

Twórcy

autor

Smyczyńska U.

• AGH University of Science and Technology, Department of Automatics and Biomedical Engineering, Cracow, Poland

autor

Smyczyńska J.

• Department of Endocrinology and Metabolic Diseases, Polish Mother’s Memorial Hospital-Research Institute in Lodz, Lodz, Poland

autor

Tadeusiewicz R.

• AGH University of Science and Technology, Department of Automatics and Biomedical Engineering, Cracow, Poland

Bibliografia

• 1. Czekalski P, Lyp K. Neural network structure optimization in pattern recognition. Stud Inform 2014;35:17–32.
• 2. Sug H. The effect of training set size for the performance of neural networks of classification. WSEAS Trans Comput 2010;9:1297–306.
• 3. Ellis D, Morgan N. Size matters: an empirical study of neural network training for large vocabulary continuous speech recognition. In: 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing, 15–19 March 1999, vol. 2. Phoenix, AZ, USA: IEEE, 1999:1013–6. DOI: 10.1109/ICASSP.1999.759875.
• 4. Reitermanova Z. Feedforward neural networks – architecture optimization and knowledge extraction. In: 17th Annual Conference of Doctoral Students-WDS 2008, 3–6 June 2008, vol. 1. Prague: Matfyzpress, 2008:159–64.
• 5. Shahamiri SR, Binti Salim SS. Real-time frequency-based noiserobust automatic speech recognition using multi-nets artificial neural networks: a multi-views multi-learners approach. Neurocomputing 2014;129:199–207.
• 6. Tamura S, Tateishi M. Capabilities of a four-layered feedforward neural network: four layers versus three. IEEE Trans Neural Networks 1997;8:251–5.
• 7. Jinchuan K, Xinzhe L. Empirical analysis of optimal hidden neurons in neural network modeling for stock prediction. In: Proc. 2008 Pacific-Asia Work Comput Intell Ind Appl PACIIA ‘08, 19–20 December 2008, vol. 2. Wuhan, China: IEEE, 2008:828–32.
• 8. Karsoliya S. Approximating number of hidden layer neurons in multiple hidden layer BPNN architecture. Int J Eng Trends Technol 2012;3:714–7.
• 9. Sheela KG, Deepa SN. Review on methods to fix number of hidden neurons in neural networks. Math Probl Eng 2013. Article ID 425740, 11 pages. DOI: 10.1155/2013/425740.
• 10. Greulich WW, Pyle SI. Radiographic atlas of skeletal development of the hand and wrist, 2nd ed. Stanford, CA, USA: Stanford University Press, 1993.
• 11. Palczewska I, Niedzwiecka Z. Somatic development indices in children and youth of Warsaw (in Polish). Dev Period Med 2001;5:18–118.
• 12. Smyczyńska U, Smyczyńska J, Hilczer M, Stawerska R, Lewiński A, Tadeusiewicz R. Artificial neural networks – a novel tool in modelling the effectiveness of growth hormone (GH) therapy in children with GH deficiency. Pediatr Endocrinol 2015;14:9–18.
• 13. Smyczyńska U, Smyczyńska J, Tadeusiewicz R. Neural modelling of growth hormone therapy for the prediction of therapy results. Bio-Algorithms Med-Syst 2015;11:33–45.
• 14. Ranke MB, Lindberg A, Chatelain P, Wilton P, Cutfield W, Albertsson-Wikland K, et al. Derivation and validation of a mathematical model for predicting the response to exogenous recombinant human growth hormone (GH) in prepubertal children with idiopathic GH deficiency. J Clin Endocrinol Metab 1999;84:1174–83.
• 15. De Ridder MA, Stijnen T, Hokken-Koelega AC. Prediction of adult height in growth-hormone-treated children with growth hormone deficiency. J Clin Endocrinol Metab 2007;92:925–31. 16. Smyczynska J, Hilczer M, Smyczynska U, Stawerska R, Tadeusiewicz R, Lewinski A. Neural network models – a novel tool for predicting the efficacy of growth hormone (GH) therapy in children with short stature. Neuroendocrinol Lett 2015;36:348–53.
• 17. Hunter D, Yu H, Pukish MS, Kolbusz J, Wilamowski BM. Selection of proper neural network sizes and architectures – a comparative study. IEEE Trans Ind Inform 2012;8:228–40.
• 18. Lawrence S, Giles CL, Tsoi AC. Lessons in neural network training: overfitting may be harder than expected. In: The Fourteenth National Conference on Artificial Intelligence, 27–31 July 1997. Providence, RI, USA: AAAI Press, 1997:540–5.

Uwagi

Opracowanie ze środków MNiSW w ramach umowy 812/P-DUN/2016 na działalność upowszechniającą naukę.