Predicting Equivalent Temperature Index in Dairy Cows with the NeuralProphet Model

Precision Livestock Farming (PLF) tools have become an integral part of sustainable livestock management, offering increased productivity while minimizing environmental impact. This study explores the application of the NeuralProphet model to predict Equivalent Temperature Index (ETI) in dairy cows, leveraging the development of powerful statistical algorithms and advances in artificial intelligence, particularly in the deep learning subset. The data used come from a smart monitoring system implemented in a pilot dairy farm located in Budrio (Emilia-Romagna, Italy) while the data of the main external environmental parameters were obtained from the weather service provider Open-Meteo, taking advantage of the performance of predictive models already tested in various applications. In this study, a predictive model for ETI was developed through an initial data exploration that includes preprocessing tasks such as outlier detection, unit adjustment, and calculation of the target variable. Exogenous variables were analyzed for sensitivity, resulting in the selection of 24-hour forecasts for outdoor temperature, relative humidity, solar radiation, and wind speed as model inputs. In addition, values for indoor temperature, relative humidity, and wind speed were included as initial conditions for model reproducibility in future studies. As a final step, the evaluation of the accuracy of the model was performed by comparing the indoor environmental parameters of the building provided by the model with raw measurements. As a result, a reliable tool was synthesized to support the decision-making process of farmers to improve livestock welfare and increase animal productivity.