COMBINING MACHINE LEARNING AND PATH ANALYSIS TO IDENTIFY ADAPTIVE MECHANISMS IN FARM ANIMALS

Abstract

Machine learning is a field of artificial intelligence which has been gaining prominence, mainly in order to solve multifactorial problems such as animal adaptation, which involves thermoregulatory, morphological, hematological, immunological and productive responses. For the first time, it proposes to trace the adaptive profile, identify biomarkers of animal adaptation and investigate the potential effects underlying the complex relationships of animal adaptation with climatic, morphological, thermoregulatory, hematological, immunological and productive variables in animal modeling using predictive machine learning as an auxiliary method. The database contains 7608 pieces of information. All variables were measured simultaneously in each animal. Machine learning techniques and path analysis were used. The objective of this proposal is to provide the development of statistical skills, specifically systematic methodologies using different supervised and unsupervised machine learning techniques to analyze big data relating to adaptation and animal production, in addition to discussing structuring, planning and data analysis using artificial intelligence with Professor Jérôme Bindelle (h-index=29), who will train FAPESP doctoral student Robson Silveira for five months to develop these skills. Finally, I emphasize that this study is part of a regular project supported by FAPESP in the area of animal ambience and machine learning associated with the development of products, services and innovation, made possible by technologies aimed at agriculture 4.0. The proposal is expected to contribute to the expansion of scientific frontiers in the sense of revealing complex processes that help to understand the adaptive responses of livestock to climate change, in addition to identifying phenotypic markers that can be used to quantify responses to heat stress.