Publicação
LSTM Models to Support the Selective Antibiotic Treatment Strategy of Dairy Cows in the Dry Period
| datacite.subject.fos | Ciências Naturais::Ciências da Computação e da Informação | pt_PT |
| datacite.subject.fos | Ciências Médicas::Ciências da Saúde | pt_PT |
| dc.contributor.advisor | Henriques, Roberto André Pereira | |
| dc.contributor.advisor | Nunes, Telmo Pina | |
| dc.contributor.author | Rafael, Joana Sofia Silva Baptista | |
| dc.date.accessioned | 2023-02-03T09:53:31Z | |
| dc.date.available | 2023-02-03T09:53:31Z | |
| dc.date.issued | 2023-01-23 | |
| dc.description | Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science | pt_PT |
| dc.description.abstract | Udder inflammation, known as mastitis, is the most significant disease of dairy cows worldwide, invoking substantial economic losses. The current common strategy to reduce this problem is the prophylactic administration of antibiotics treatment of cows during their dry period. Paradoxically, the indiscriminate use of antibiotics in animals and humans has been the leading cause of antimicrobial resistance, a concern in several public health organizations. In light of these assumptions, at the beginning of 2022, the European Union made it illegal to routinely administer antibiotics on farms, with Regulation 2019/6 of 11 December 2018. Considering this new scenario, the objective of this study was to produce a model that supports the decisions of veterinarians when administering antibiotics in the dry period of dairy cows. Deep learning models were used, namely LSTM layers that operate with dynamic features from milk recordings and a dense layer that uses static features. Two approaches were chosen to deal with this problem. The first is based on a binary classification model that considers the occurrence of mastitis within 60 days after calving. The second approach was a multiclass classification model based on veterinary expert judgment. In each approach, three models were implemented, a Vanilla LSTM, a Stacked LSTM, and a Stacked LSTM with a dense layer working in parallel. The best performances from binary and multiclass approaches were 65% and 84% accuracy, respectively. It was possible to conclude that the models of the multiclass classification approach had better performance than the other classification. The capture of long- and short-term dependencies in the LSTM models, especially with the combination of static features, obtained promising results, which will undoubtedly contribute to producing a machine learning system with a prompt and affordable response, allowing for a reduction in the administration of antibiotics in dairy cows to the strictly necessary. | pt_PT |
| dc.identifier.tid | 203212622 | pt_PT |
| dc.identifier.uri | http://hdl.handle.net/10362/148602 | |
| dc.language.iso | eng | pt_PT |
| dc.rights.uri | http://creativecommons.org/licenses/by-nd/4.0/ | pt_PT |
| dc.subject | Machine Learning | pt_PT |
| dc.subject | Deep Learning | pt_PT |
| dc.subject | LSTM | pt_PT |
| dc.subject | Dairy Cows | pt_PT |
| dc.subject | Antimicrobial Resistance | pt_PT |
| dc.subject | Mastitis | pt_PT |
| dc.subject | SDG 15 - Life on land | |
| dc.subject | SDG 12 - Responsible production and consumption | |
| dc.subject | SDG 3 - Good health and well-being | |
| dc.title | LSTM Models to Support the Selective Antibiotic Treatment Strategy of Dairy Cows in the Dry Period | pt_PT |
| dc.type | master thesis | |
| dspace.entity.type | Publication | |
| rcaap.rights | openAccess | pt_PT |
| rcaap.type | masterThesis | pt_PT |
| thesis.degree.name | Mestrado em Ciência de Dados e Métodos Analíticos Avançados, especialização em Ciência de Dados | pt_PT |