東華大學圖書館 |

Machine Learning in Poultry Companies' Data. Applications and Methodologies.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Machine Learning in Poultry Companies' Data. Applications and Methodologies./
Author:	de Carvalho, Luiz Victor Ferrato Melo.
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2022,
Description:	162 p.
Notes:	Source: Dissertations Abstracts International, Volume: 83-11, Section: B.
Contained By:	Dissertations Abstracts International83-11B.
Subject:	Internet of Things. -
ISBN:	9798426868779

Machine Learning in Poultry Companies' Data. Applications and Methodologies.
de Carvalho, Luiz Victor Ferrato Melo.

Machine Learning in Poultry Companies' Data. Applications and Methodologies. - Ann Arbor : ProQuest Dissertations & Theses, 2022 - 162 p.

Source: Dissertations Abstracts International, Volume: 83-11, Section: B.

Thesis (Ph.D.)--North Carolina State University, 2022.

This item must not be sold to any third party vendors.

All agricultural sectors must seek efficiency improvements to meet the growing demand for food based on the expected population increase over the next few decades. Poultry products play an essential role in meeting global protein demand due to its their production efficiency, quality, affordable prices, and few cultural restrictions. To further improve the efficiency of this sector, all available tools must be explored, including computational tools that help manage and analyze large datasets to identify factors of influence and decision-making in animal production. However, the application of computational tools, such as Machine Learning (ML), using data already collected by commercial poultry production companies have been published. The hypothesis of this work is ML Algorithms can be used for predictive modeling and decisionmaking from production data routinely collected by commercial broiler production companies. This dissertation aims to demonstrate how to organize common production data, extract more information from this data that would not be possible by statistical methods, and provide methodologies of ML applications to benefit academic and industrial research.The methodologies developed in this work were based on three significant applications. The first aims to identify variables of interest to direct research and investments. The second methodology can create tailored models for application in industry and research. The last one is to develop decision-making tools that can be used to support feed additive decisions and be used as a sales tool for the animal feed industry. To develop the methodologies and applications, five Brazillian poultry production companies agreed to share their production data from 2016 until 2021. The data is proprietary, so the data and models created from it are not available to be published or shared. Nevertheless, the outcome of this work is the methodologies and the codes, which are available in full in this dissertation. A total of 1,354 files in different formats were received with information of 7,188 flocks, 661 feed formulations, and 274 broiler grow-out houses. Chapter 1 describes the organization of the data and the data augmentation process used, which increased the number of variables from 45 to 846. As a result, a 23.3 MB Comma Separated Value (CSV) file containing 7,188 instances (poultry flocks) and 846 attributes (variables) was generated. Chapter 2 describes and demonstrates a ML methodology utilizing the FeatureWiz feature selection library, capable of identifying variables of interest to a target factor. To demonstrate the methodology, FCR was set as the target, and as a result, the method highlighted outside weather significant factors affecting FCR what could indicate, among other possibilities, a problem on the isolation of the houses; or lead to new research topics associated with animal behavior influenced by the weather. Chapter 3 utilizes H2O AutoML and AutoKeras, combined with FeatureWiz, to develop a method capable of creating tailored models based on companies' data. The methodology is applied and demonstrated having as target the Real Feed Intake (RFI), which was calculated based on a Lysine requirement factorial model. The validation of the method was achieved by the correct identification of the variables used to calculate RFI and the ability to identify and replicate the results with a low RMSE (0.0629) as compared to the values generated with the lysine model. At the end of this chapter, an adaptation of the methodology is explained and demonstrated to create decision-making tools. In conclusion, this work demonstrated that data commonly collected by poultry production companies can be combined with ML for modeling and decision-making tools creation and provides, for the proposed outcomes, the complete methodologies and codes that are simple and free to use yet powerful.

ISBN: 9798426868779Subjects--Topical Terms:

3538511
Internet of Things.

Machine Learning in Poultry Companies' Data. Applications and Methodologies.
LDR:04859nmm a2200289 4500 001 2349725
005 20221003074939.5
008 241004s2022 eng d
020 $a 9798426868779
035 $a (MiAaPQ)AAI29071243
035 $a (MiAaPQ)NCState_Univ18402039362
035 $a AAI29071243
040 $a MiAaPQ $c MiAaPQ
100 1 $a de Carvalho, Luiz Victor Ferrato Melo. $3 3689144
245 1 0 $a Machine Learning in Poultry Companies' Data. Applications and Methodologies.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2022
300 $a 162 p.
500 $a Source: Dissertations Abstracts International, Volume: 83-11, Section: B.
500 $a Advisor: Malheiros, Ramon;Ferket, Peter.
502 $a Thesis (Ph.D.)--North Carolina State University, 2022.
506 $a This item must not be sold to any third party vendors.
520 $a All agricultural sectors must seek efficiency improvements to meet the growing demand for food based on the expected population increase over the next few decades. Poultry products play an essential role in meeting global protein demand due to its their production efficiency, quality, affordable prices, and few cultural restrictions. To further improve the efficiency of this sector, all available tools must be explored, including computational tools that help manage and analyze large datasets to identify factors of influence and decision-making in animal production. However, the application of computational tools, such as Machine Learning (ML), using data already collected by commercial poultry production companies have been published. The hypothesis of this work is ML Algorithms can be used for predictive modeling and decisionmaking from production data routinely collected by commercial broiler production companies. This dissertation aims to demonstrate how to organize common production data, extract more information from this data that would not be possible by statistical methods, and provide methodologies of ML applications to benefit academic and industrial research.The methodologies developed in this work were based on three significant applications. The first aims to identify variables of interest to direct research and investments. The second methodology can create tailored models for application in industry and research. The last one is to develop decision-making tools that can be used to support feed additive decisions and be used as a sales tool for the animal feed industry. To develop the methodologies and applications, five Brazillian poultry production companies agreed to share their production data from 2016 until 2021. The data is proprietary, so the data and models created from it are not available to be published or shared. Nevertheless, the outcome of this work is the methodologies and the codes, which are available in full in this dissertation. A total of 1,354 files in different formats were received with information of 7,188 flocks, 661 feed formulations, and 274 broiler grow-out houses. Chapter 1 describes the organization of the data and the data augmentation process used, which increased the number of variables from 45 to 846. As a result, a 23.3 MB Comma Separated Value (CSV) file containing 7,188 instances (poultry flocks) and 846 attributes (variables) was generated. Chapter 2 describes and demonstrates a ML methodology utilizing the FeatureWiz feature selection library, capable of identifying variables of interest to a target factor. To demonstrate the methodology, FCR was set as the target, and as a result, the method highlighted outside weather significant factors affecting FCR what could indicate, among other possibilities, a problem on the isolation of the houses; or lead to new research topics associated with animal behavior influenced by the weather. Chapter 3 utilizes H2O AutoML and AutoKeras, combined with FeatureWiz, to develop a method capable of creating tailored models based on companies' data. The methodology is applied and demonstrated having as target the Real Feed Intake (RFI), which was calculated based on a Lysine requirement factorial model. The validation of the method was achieved by the correct identification of the variables used to calculate RFI and the ability to identify and replicate the results with a low RMSE (0.0629) as compared to the values generated with the lysine model. At the end of this chapter, an adaptation of the methodology is explained and demonstrated to create decision-making tools. In conclusion, this work demonstrated that data commonly collected by poultry production companies can be combined with ML for modeling and decision-making tools creation and provides, for the proposed outcomes, the complete methodologies and codes that are simple and free to use yet powerful.
590 $a School code: 0155.
650 4 $a Internet of Things. $3 3538511
650 4 $a Artificial intelligence. $3 516317
650 4 $a Decision making. $3 517204
650 4 $a Neural networks. $3 677449
690 $a 0800
710 2 0 $a North Carolina State University. $3 1018772
773 0 $t Dissertations Abstracts International $g 83-11B.
790 $a 0155
791 $a Ph.D.
792 $a 2022
793 $a English