ABSTRACT The objective of this study was to assess behavior patterns in Brazilian farm milk prices. We employed a structural time series techniques model, the Unobserved Component Model (UCM), which is part of the family of State Space models, to assess the trend, seasonality, cyclical behavior, and impacts of exogenous regressors on aggregated farm milk price behavior in Brazil from January 2005 to December 2019. We tested five alternative models with different regressors using the monthly national average prices of milk paid to farmers. The fit of the models was assessed with Akaike information criterion and Bayesian information criterion. Predictions were assessed by the root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). All models demonstrated a high degree of accuracy. Trends, seasonality, and two cycles were statistically significant, with the trend and long-period cycle contributing the most to price variation. Exogenous factors such as feed cost and international dairy product prices also had significant positive effects on the level of Brazil’s farm milk prices. All models demonstrated a high degree of accuracy, which may indicate their usefulness for price forecasting and policy formulation. model UCM, UCM , (UCM) seasonality 200 2019 farmers RMSE, RMSE (RMSE) MAE, MAE (MAE) MAPE. MAPE . (MAPE) accuracy Trends longperiod long period variation Brazils s formulation (UCM 20 201 (RMSE (MAE (MAPE 2

Resumo O objetivo desse trabalho é fazer a análise climática, utilizando modelos de redes neurais artificiais, como auxílio no processo de tomada de decisão no plantio de alguns tipos de produtos agrícolas. Um banco de dados com os principais elementos do clima foi construído a partir do site do INMET (Instituto Nacional de Meteorologia) e desses elementos foram identificados, em um nível de significância de 0,05, os que mais influenciavam no valor da temperatura média. Modelos de Redes Neurais Artificiais foram criados e avaliados por meio do MAD (Mean Absolute Deviation), MSE (Mean Squared Error), RMSE (Root-Mean Squared Error) e MAPE (Mean Absolute Percentage Error), e finalmente, associou-se o valor da previsão com o melhor cultivo agrícola a ser aplicado. Foram elaboradas 12 redes neurais, oito são referentes à previsão da temperatura e as outras quatro à previsão da precipitação. As redes que demostraram melhor desempenho são as que consideram todos os elementos do clima. É possível concluir que as redes neurais demostraram um desempenho adequado na previsão de séries temporais caóticas, e consequentemente, tiveram o seu resultado associado com o melhor cultivo a ser aplicado para cada previsão. Um cronograma é apresentado no final, sugerindo a melhor época para plantar cada uma das culturas estudadas. Observa-se que para os próximos cinco anos, o cultivo mais adequado à faixa prevista é o da cenoura.

Abstract The aim of this study is to conduct climate forecasting with models of artificial neural networks as a tool in the decision-making process for the planting of some types of agricultural products. A database with the main climate elements was built from the National Institute of Meteorology (INMET), and those elements that influenced the average temperature value the most were found at a significance level of 0.05. Models of Artificial Neural Networks were developed and tested using Mean Absolute Deviation (MAD), Mean Squared Error (MSE), Root-Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE), before being linked to the best agricultural cultivation forecast value. Twelve neural networks were elaborated, eight of them are related to the temperature forecast and the other four are related to the precipitation forecast. The networks that showed the best performance are those that consider all the elements of climate. It is possible to conclude that the artificial neural networks showed an adequate performance in predicting chaotic time series, and that their results were therefore linked to the optimum cultivation to use for each forecast. A schedule is supplied at the end, indicating the ideal time to plant each of the crops evaluated. Carrot is found to be the best suited crop for the forecasted range over the next five years.

ABSTRACT This study aimed to evaluate the lactation curve of female Murrah buffaloes, using mixed nonlinear models (NLM), across three lactation periods (180 d, 210 d, and 240 d). A total of 5334 data on daily milk production (kg) were analyzed. The data were collected every seven days in the interval of one to 250 days of lactation, corresponding to 221 lactations and 145 females, with calvings from 2017 to 2019. The data came from a herd located in the Centro municipality, Tabasco, Mexico. Five NLM were evaluated: Wood (WOD), Wiltmink (WIL), Cobby (COB), Brody (BRO), Sikka (SIK). The best fit model was selected based on the mean prediction error, mean absolute percentage error (MAPE), prediction error variance, coefficient of determination (R2), concordance correlation coefficient (CCC), Akaike (AIC) and Bayesian (BIC) information criteria. A regression analysis was performed between the observed and predicted values. All the NLM had a R2 above 0.91. They tend to underestimate the predictions, without residual autocorrelation. The MAPE showed an average value of 23.5%. The best fit model was WOD, followed by SIK and BRO. For WIL and COB, the mixed model did not improve the fitting. The shortest lactation period showed the best fit, followed by the 210 d and 240 d periods. The relationship between observed:predicted values fluctuated from 0.65 to 1.00, with an average value of 0.94. The use of NLM transcended in the AIC and BIC. The evaluated models showed goodness of fit, with good predictability, but low values in accuracy and precision of prediction.