Data Driven Broiler Weight Forecasting using Dynamic Neural Network Models

Simon Vestergaard Johansen, Jan Dimon Bendtsen, Martin Riisgaard-Jensen, Jesper Mogensen

Research output: Contribution to journalConference article in JournalResearchpeer-review

3 Citations (Scopus)
146 Downloads (Pure)

Abstract

In this article, the dynamic influence of environmental broiler house conditions and broiler growth is investigated.
Dynamic neural network forecasting models have been trained on farm-scale broiler batch production data from 12 batches from the same house.
The model forecasts future broiler weight and uses environmental conditions such as heating, ventilation, and temperature along with broiler behavior such as feed and water consumption.
Training data and forecasting data is analyzed to explain when the model might fail at generalizing.
We present ensemble broiler weight forecasts to day 7, 14, 21, 28 and 34 from all preceding days and provide our interpretation of the results.
Results indicate that the dynamic interconnection between environmental conditions and broiler growth can be captured by the model.
Furthermore, we found that a comparable forecast can be obtained by using input data from the previous batch as a substitute for future input data.
Original languageEnglish
Book seriesIFAC-PapersOnLine
Volume50
Issue number1
Pages (from-to)5398-5403
ISSN2405-8963
DOIs
Publication statusPublished - 8 Jul 2017
Event2017 IFAC Congress -
Duration: 9 Jul 201714 Jul 2017
https://www.ifac2017.org/

Conference

Conference2017 IFAC Congress
Period09/07/201714/07/2017
Internet address

Fingerprint

Neural networks
Farms
Ventilation
Heating
Water
Temperature

Keywords

  • Artificial neural nets in agriculture

Cite this

Johansen, Simon Vestergaard ; Bendtsen, Jan Dimon ; Riisgaard-Jensen, Martin ; Mogensen, Jesper. / Data Driven Broiler Weight Forecasting using Dynamic Neural Network Models. In: IFAC-PapersOnLine. 2017 ; Vol. 50, No. 1. pp. 5398-5403.
@inproceedings{a893363e3ec94169b58be37fa3d791ba,
title = "Data Driven Broiler Weight Forecasting using Dynamic Neural Network Models",
abstract = "In this article, the dynamic influence of environmental broiler house conditions and broiler growth is investigated.Dynamic neural network forecasting models have been trained on farm-scale broiler batch production data from 12 batches from the same house.The model forecasts future broiler weight and uses environmental conditions such as heating, ventilation, and temperature along with broiler behavior such as feed and water consumption. Training data and forecasting data is analyzed to explain when the model might fail at generalizing.We present ensemble broiler weight forecasts to day 7, 14, 21, 28 and 34 from all preceding days and provide our interpretation of the results. Results indicate that the dynamic interconnection between environmental conditions and broiler growth can be captured by the model.Furthermore, we found that a comparable forecast can be obtained by using input data from the previous batch as a substitute for future input data.",
keywords = "Artificial neural nets in agriculture",
author = "Johansen, {Simon Vestergaard} and Bendtsen, {Jan Dimon} and Martin Riisgaard-Jensen and Jesper Mogensen",
year = "2017",
month = "7",
day = "8",
doi = "10.1016/j.ifacol.2017.08.1073",
language = "English",
volume = "50",
pages = "5398--5403",
journal = "I F A C Workshop Series",
issn = "1474-6670",
publisher = "Elsevier",
number = "1",

}

Data Driven Broiler Weight Forecasting using Dynamic Neural Network Models. / Johansen, Simon Vestergaard; Bendtsen, Jan Dimon; Riisgaard-Jensen, Martin; Mogensen, Jesper.

In: IFAC-PapersOnLine, Vol. 50, No. 1, 08.07.2017, p. 5398-5403.

Research output: Contribution to journalConference article in JournalResearchpeer-review

TY - GEN

T1 - Data Driven Broiler Weight Forecasting using Dynamic Neural Network Models

AU - Johansen, Simon Vestergaard

AU - Bendtsen, Jan Dimon

AU - Riisgaard-Jensen, Martin

AU - Mogensen, Jesper

PY - 2017/7/8

Y1 - 2017/7/8

N2 - In this article, the dynamic influence of environmental broiler house conditions and broiler growth is investigated.Dynamic neural network forecasting models have been trained on farm-scale broiler batch production data from 12 batches from the same house.The model forecasts future broiler weight and uses environmental conditions such as heating, ventilation, and temperature along with broiler behavior such as feed and water consumption. Training data and forecasting data is analyzed to explain when the model might fail at generalizing.We present ensemble broiler weight forecasts to day 7, 14, 21, 28 and 34 from all preceding days and provide our interpretation of the results. Results indicate that the dynamic interconnection between environmental conditions and broiler growth can be captured by the model.Furthermore, we found that a comparable forecast can be obtained by using input data from the previous batch as a substitute for future input data.

AB - In this article, the dynamic influence of environmental broiler house conditions and broiler growth is investigated.Dynamic neural network forecasting models have been trained on farm-scale broiler batch production data from 12 batches from the same house.The model forecasts future broiler weight and uses environmental conditions such as heating, ventilation, and temperature along with broiler behavior such as feed and water consumption. Training data and forecasting data is analyzed to explain when the model might fail at generalizing.We present ensemble broiler weight forecasts to day 7, 14, 21, 28 and 34 from all preceding days and provide our interpretation of the results. Results indicate that the dynamic interconnection between environmental conditions and broiler growth can be captured by the model.Furthermore, we found that a comparable forecast can be obtained by using input data from the previous batch as a substitute for future input data.

KW - Artificial neural nets in agriculture

U2 - 10.1016/j.ifacol.2017.08.1073

DO - 10.1016/j.ifacol.2017.08.1073

M3 - Conference article in Journal

VL - 50

SP - 5398

EP - 5403

JO - I F A C Workshop Series

JF - I F A C Workshop Series

SN - 1474-6670

IS - 1

ER -