Abstract
A Neural Network Model to predict the urinary excretion of purine derivative nitrogen (UPDN) in cows is presented. The input variables of the model are dry matter intake (DMINT), NDF intake (NDFINT), total soluble nitrogen (SP), total soluble non-protein dry matter (SNPDM), total degradable nitrogen (DCP), total degradable non-protein dry matter (DNPDM), hourly available CP in the rumen (HACP), hourly available non-protein dry matter (HANPDM), three different gross indexes of synchronization, namely SYNCA (SP/SNPDM), SYNCB (DCP/DNPDM) and SYNCK (HACP/HANPDM) and two variables describing some metabolic aspects of purine derivative excretion such as live weight of the cow (LW) and milk yield (MILKY). The Model developed uses the Multi Layer Perceptron (MLP) utility, with 13 nodes in the input layer, 8 nodes in the hidden layer and 1 node in the output layer. The Model performances have been tested over 24 observations not previously used to train the model. When compared to a linear regression approach, the Neural Network model showed better performance but under predicted the daily excretion of UPDN for values around 20 g/day. When evaluated in terms of behaviour and depicted scenario the model responded to changes of live weight (LW) and milk yield (MILKY) and to modifications
More>>
Volpe, V;
Stefanon, B;
Moscardini, S;
Susmel, P;
[1]
Gruber, L
[2]
- University of Udine, Department of Animal Production Science, Pagnacco, UD (Italy)
- Federal Research Institute for Agriculture in the Alpine Regions, Irdning (Austria)
Citation Formats
Volpe, V, Stefanon, B, Moscardini, S, Susmel, P, and Gruber, L.
Development of a neutral network model to predict the excretion of purine derivatives in the urine of cows.
IAEA: N. p.,
1999.
Web.
Volpe, V, Stefanon, B, Moscardini, S, Susmel, P, & Gruber, L.
Development of a neutral network model to predict the excretion of purine derivatives in the urine of cows.
IAEA.
Volpe, V, Stefanon, B, Moscardini, S, Susmel, P, and Gruber, L.
1999.
"Development of a neutral network model to predict the excretion of purine derivatives in the urine of cows."
IAEA.
@misc{etde_363475,
title = {Development of a neutral network model to predict the excretion of purine derivatives in the urine of cows}
author = {Volpe, V, Stefanon, B, Moscardini, S, Susmel, P, and Gruber, L}
abstractNote = {A Neural Network Model to predict the urinary excretion of purine derivative nitrogen (UPDN) in cows is presented. The input variables of the model are dry matter intake (DMINT), NDF intake (NDFINT), total soluble nitrogen (SP), total soluble non-protein dry matter (SNPDM), total degradable nitrogen (DCP), total degradable non-protein dry matter (DNPDM), hourly available CP in the rumen (HACP), hourly available non-protein dry matter (HANPDM), three different gross indexes of synchronization, namely SYNCA (SP/SNPDM), SYNCB (DCP/DNPDM) and SYNCK (HACP/HANPDM) and two variables describing some metabolic aspects of purine derivative excretion such as live weight of the cow (LW) and milk yield (MILKY). The Model developed uses the Multi Layer Perceptron (MLP) utility, with 13 nodes in the input layer, 8 nodes in the hidden layer and 1 node in the output layer. The Model performances have been tested over 24 observations not previously used to train the model. When compared to a linear regression approach, the Neural Network model showed better performance but under predicted the daily excretion of UPDN for values around 20 g/day. When evaluated in terms of behaviour and depicted scenario the model responded to changes of live weight (LW) and milk yield (MILKY) and to modifications of the pattern of nutrients supplied to rumen microbes. (author) 16 refs, 11 figs, 1 tab}
place = {IAEA}
year = {1999}
month = {Jun}
}
title = {Development of a neutral network model to predict the excretion of purine derivatives in the urine of cows}
author = {Volpe, V, Stefanon, B, Moscardini, S, Susmel, P, and Gruber, L}
abstractNote = {A Neural Network Model to predict the urinary excretion of purine derivative nitrogen (UPDN) in cows is presented. The input variables of the model are dry matter intake (DMINT), NDF intake (NDFINT), total soluble nitrogen (SP), total soluble non-protein dry matter (SNPDM), total degradable nitrogen (DCP), total degradable non-protein dry matter (DNPDM), hourly available CP in the rumen (HACP), hourly available non-protein dry matter (HANPDM), three different gross indexes of synchronization, namely SYNCA (SP/SNPDM), SYNCB (DCP/DNPDM) and SYNCK (HACP/HANPDM) and two variables describing some metabolic aspects of purine derivative excretion such as live weight of the cow (LW) and milk yield (MILKY). The Model developed uses the Multi Layer Perceptron (MLP) utility, with 13 nodes in the input layer, 8 nodes in the hidden layer and 1 node in the output layer. The Model performances have been tested over 24 observations not previously used to train the model. When compared to a linear regression approach, the Neural Network model showed better performance but under predicted the daily excretion of UPDN for values around 20 g/day. When evaluated in terms of behaviour and depicted scenario the model responded to changes of live weight (LW) and milk yield (MILKY) and to modifications of the pattern of nutrients supplied to rumen microbes. (author) 16 refs, 11 figs, 1 tab}
place = {IAEA}
year = {1999}
month = {Jun}
}