Using neural networks to prediction of bulls' hides technological traits

To chci

The aim of study was adaptation of the artificial neural network to predict some of post-slaughter bulls hide parameters based on their body conformation. Material for the experiment was 105 bulls. Four variants of neural network were adopted which were different in length of learning (in 500, 1000, 1200 and 1500 epochs). The best variant appeared which was learning in 1200 epochs. Therefore this network's variant was precisely evaluated. High conformity expressed by a coefficient of correlation has been proved for the following traits: surface of raw hide (r=0,86) and for a leather (r=0,88) and also for a weight of raw hide (r=0,81). However the neural network was ineffective in designation difference between a thickness in a butt leather standard point and a thickness in a point of fall of a thickness at shoulder blade region, because of a big prediction error which could be reduce by modification of network model and by increase number of the observations which took part in network learning process. … celý popis

Podrobná bibliografie

Další autoři:

Krzysztof Adamczyk, Piotr Zapletal, Andrzej Węglarz, Grzegorz Skrzyński, Katarzyna Kuza, Václav Řehout, 1943- , Jan Frelich, 1941- , Jarmila Voříšková, 1956-

Typ dokumentu:

Články

Publikováno v:

Collection of Scientific Papers, Faculty of Agriculture in České Budějovice. Series for Animal Sciences. -- ISSN 1212-558X. -- Roč. 22, č. 1 (2005), s. 91-99

Témata:

skot

býci

usně

zpracovatelská kvalita

technické vlastnosti

rozměry

plocha povrchu

hmotnost

prognózování

matematické modelování

počítače

CATTLE

BULLS

HIDES AND SKINS

PROCESSING QUALITY

TECHNICAL PROPERTIES

DIMENSIONS

SURFACE AREA

WEIGHT

FORECASTING

MATHEMATICAL MODELS

COMPUTERS

umělá neuronová síť

Popis jednotky:

Grafy, tab.

Fyzický popis:

11 il.

Bibliografie:

Lit. na s. 99