HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Ahmadi, H. Articles by Nariman-Zadeh, N. Search for Related Content PubMed PubMed Citation Articles by Ahmadi, H. Articles by Nariman-Zadeh, N.

Prediction Model for True Metabolizable Energy of Feather Meal and Poultry Offal Meal Using Group Method of Data Handling-Type Neural Network

H. Ahmadi^*,1, A. Golian^*, M. Mottaghitalab and N. Nariman-Zadeh

^* Center of Excellence in the Animal Science Department, Ferdowsi University of Mashhad, Mashhad, Iran 91775-1163; Department of Animal Science, University of Guilan, Rasht, Iran 41635-1314; and Department of Mechanical Engineering, University of Guilan, Rasht, Iran 41635-3756

¹ Corresponding author: hahmadima{at}yahoo.com

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 
A group method of data handling-type neural network (GMDH-type NN) with an evolutionary method of genetic algorithm was used to predict the TME_n of feather meal (FM) and poultry offal meal (POM) based on their CP, ether extract, and ash content. Thirty-seven data lines consisting of 15 FM and 22 POM samples were collected from literature and used to train a GMDH-type NN model. A genetic algorithm was deployed to design the whole architecture of the GMDH-type NN. The accuracy of the model was examined by R² value, adjusted R², mean square error, residual standard deviation, mean absolute percentage error, and bias. The developed model could accurately predict the TME_n of FM or POM samples from their chemical composition. The R² for the GMDH-type NN model had a higher accuracy of prediction than 2 models reported previously. This study revealed that the novel modeling of GMDH-type NN with method of genetic algorithm can be used to predict the TME_n of poultry by-products.

Key Words: feather meal • poultry offal meal • metabolizable energy • neural network model

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 
Once the nutrient requirements of the animals were established, a correct balanced diet can be formulated if the accurate nutrient composition of feedstuffs is known. The feather meal (FM) and poultry offal meal (POM) are widely used in broiler diets, and the accurate information on their energy content are of importance to renderers and the nutritionists (Dale and Batal, 2002). It is reported that the energy content of feedstuffs, especially FM and POM in terms of TME_n, strongly depends on their chemical composition. The nutritionists are interested in using a model that can precisely predict the TME_n value of animal by-products. The researchers have tried to develop a TME_n prediction model for the FM and POM samples (Dale, 1992; Dale et al., 1993). All the previously TME_n prediction models reported for poultry by-product meals were based on the regression analysis methods using their CP, ether extract (EE), and ash content. Alternatively, a soft-computing method of artificial neural network (ANN) seemed to be more appropriate for the TME_n prediction of a feedstuff. An ANN is a set of nonlinear equations that predicts output variable(s) from input variable(s) in a flexible way using layers of linear regressions and S-shaped functions. There is no need to provide an a priori model when using an ANN. It is potentially advantageous in modeling the biological processes often characterized as highly nonlinear (Dayhoff and DeLeo, 2001). The ANN functions are applied to many fields to model and predict the behaviors of unknown systems, very complex systems, or both based on given input-output values (Dayhoff and DeLeo, 2001).

One submodel of ANN is a group method of data handling-type neural network (GMDH-type NN). It is a self-organizing approach by which gradually more complex models are generated from their performance evaluation and a set of multi-input, single output data pairs (Lemke and Mueller, 2003). The GMDH was first developed by Ivakhnenko (1971) as a multivariate analysis method for modeling and identification of complex systems. The main idea of GMDH is to build an analytical function in a feed-forward network based on a quadratic node transfer function whose coefficients obtained by using a regression technique (Farlow, 1984). Recently, the use of such self-organizing networks has led to a successful application of the GMDH-type algorithm in a broad range of areas in engineering, science, and economics (Amanifard et al., 2008). In the poultry field, Ahmadi et al. (2007) demonstrated that the GM-DH-type NN model could provide an effective means of describing the data patterns in broiler performance prediction based on their dietary nutrients.

The purpose of this study was to examine the validity of GMDH-type NN with a genetic algorithm method to predict the TME_n of feather meal and poultry offal meal based on their chemical analysis.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 
Data Source

Thirty-seven raw data lines consisting of 15 FM and 22 POM samples were used to train a GMDH-type NN. The FM and POM data were those reported by Dale (1992) and Dale et al. (1993), respectively. Each data line consisted of CP, EE, and ash percentages and a measured TME_n for an individual sample (Table 1). Similar procedures were used to determine the chemical composition and the TME_n of meal samples (Dale, 1992; Dale et al., 1993).

A detailed description of a GMDH-type NN terminology, development, application, and examples of using this approach were reported by several researchers (Farlow, 1984; Mueller and Lemke, 2000; Lemke and Mueller, 2003; Nariman-Zadeh et al., 2005). The parameters of interest in this multi-input, single-output system that influenced the TME_n were CP, EE, and ash content of the samples. The raw data were divided into 2 parts of training and validation sets. Thirty input-output data lines (12 from FM and 18 from POM samples) were randomly selected and used to train the GMDH-type NN model as a training set. The validation set consisted of the 7 remaining data lines (3 from FM and 4 from POM samples), which were used to validate the prediction of the evolved neural network during the training processes. The data set was imported into a GEvoM for GMDH-type NN training (GEvoM, 2008). Two hidden layers were considered for prediction of the TME_n model. A population of 15 individual values with a crossover probability of 0.7, mutation probability of 0.07, and 300 generations was used to genetically design the neural network (Yao, 1999). It appeared that no further improvement could be achieved for this population size.

A quantitative verifying fit for the predictive model was made using error measurement indices commonly used to evaluate forecasting models. The goodness of fit or accuracy of the model was determined by R² value, adjusted R², mean square error, residual standard deviation, mean absolute percentage error, and bias (Oberstone, 1990).

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 
The optimal structure of the evolved 2-hidden-layer GMDH-type NN was produced from the genetic algorithm (GA) for modeling the TME_n was found with 4 hidden neurons. These neurons were the output of the GA to fit the GMDH-type NN model. The constructed model was characterized by a superb response for all input variables from the training data set. It appeared that the percentage of CP, EE, and ash had a strong effect on the TME_n prediction.

The corresponding polynomial equations used to develop the TME_n model were as follows:

([1])

([2])

([3])

([4])

([5])

The TME_n values predicted by the GMDH-type NN model is shown in Table 1. The observed and predicted values of TME_n from the training and validation sets are shown in Figure 1. The comparison of observed and predicted TME_n describes the behavior of the GMDH-type NN model from investigating inputs. The results revealed a very good agreement between the observed and predicted TME_n values (for training or validation). The GMDH-type NN model could accurately predict the TME_n of the validation data set that was not used during the training processes. The statistical tests (in terms of R², adjusted R², mean square error, residual standard deviation, and mean absolute percentage error) indicated that there was a relatively better prediction of TME_n for the validation as compared with the training values (Table 2). Dale (1992) proposed a linear equation to predict the TME_n of FM samples with one variable of EE (R² = 0.81). Dale et al. (1993) developed several prediction equations for estimating the TME_n of POM samples with 1 (EE), 2 (EE and ash), or 3 input (EE, ash, and CP) variables, or a combination of these. They suggested that the most accurate prediction equation was obtained with 3 input variables (R² of 0.81). We proposed neural network model with 3 variables and a combination of 30 data lines of FM and POM samples to compare the results of this study with those of Dale (1992) and Dale et al. (1993). The goodness of fit in terms of R² corresponding to GMDH-type NN model showed a higher accuracy of prediction than equations (0.96 vs. 0.81) reported by Dale (1992) and Dale et al. (1993).

View larger version (22K): [in this window] [in a new window]   Figure 1. The comparison of observed and model predicted TMEn values obtained from training (1 to 12 and 13 to 30 are feather and offal samples, respectively) and validation (31 to 33 and 34 to 37 are feather and offal samples, respectively) sets.

Received for publication December 17, 2007. Accepted for publication May 19, 2008.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS AND DISCUSSION REFERENCES

 
Ahmadi, H., M. Mottaghitalab, and N. Nariman-Zadeh. 2007. Group method of data handling-type neural network prediction of broiler performance based on dietary metabolizable energy, methionine, and lysine. J. Appl. Poult. Res. 16:494–501.[Abstract/Free Full Text]

Amanifard, N., N. Nariman-Zadeh, M. Borji, A. Khalkhali, and A. Habibdoust. 2008. Modelling and pareto optimization of heat transfer and flow coefficients in microchannels using GMDH type neural networks and genetic algorithms. Energy Convers. Manage. 2:311–325.

Dale, N. 1992. True metabolizable energy of feather meal. J. Appl. Poult. Res. 1:331–334.[Abstract/Free Full Text]

Dale, N., and A. Batal. 2002. Feedstuffs Ingredient Analysis Table. Miller Publishing Co., Minnetonka, MN.

Dale, N., B. Fancher, M. Zumbado, and A. Viuacres. 1993. Metabolizable energy content of poultry offal meal. J. Appl. Poult. Res. 2:40–42.[Abstract/Free Full Text]

Dayhoff, J. E., and J. M. DeLeo. 2001. Artificial neural networks: Opening the black box. Cancer 91(Suppl. 8):1615–1635.[CrossRef][Web of Science][Medline]

Farlow, S. J. 1984. Self-organizing methods in modeling. GMDH type algorithms. Marcel Dekker Inc., New York, NY.

GEvoM. 2008. GMDH-type neural network designed by an evolutionary method of modelling. http://research.guilan.ac.ir/gevom/ Accessed Jan. 20, 2008.

Ivakhnenko, A. G. 1971. Polynomial theory of complex systems. IEEE Trans. Syst. Man Cybern. 4:364–378.

Lemke, F., and J. A. Mueller. 2003. Medical data analysis using self-organizing data mining technologies. Syst. Anal. Model. Simul. 10:1399–1408.

Mueller, J. A., and F. Lemke. 2000. Self-organising data mining: An intelligent approach to extract knowledge from data. Hamburg Pub. Libri., Germany.

Nariman-Zadeh, N., A. Darvizeh, A. Jamali, and A. Moieni. 2005. Evolutionary design of generalized polynomial neural networks for modelling and prediction of explosive forming process. J. Mater. Process. Technol. 164–165:1561–1571.

Oberstone, J. 1990. Management Science-Concepts, Insights, and Applications. West Publ. Co., New York, NY.

Yao, X. 1999. Evolving artificial neural networks. Proc. IEEE 9:1423–1447.

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Ahmadi, H. Articles by Nariman-Zadeh, N. Search for Related Content PubMed PubMed Citation Articles by Ahmadi, H. Articles by Nariman-Zadeh, N.