Advanced milk production modelling using high-order generalized least deviation method

The United States' agricultural sector depends heavily on production to continuously supply the country's enormous demand for milk and dairy products. Despite its significance, long-term research on forecasting milk production volume has been limited. This study addresses this gap by analyzing trends and predicting monthly milk production volume using the Generalized Least Deviation Method (GLDM). A comprehensive dataset of 168 months of U.S. milk production data was utilized. Various-order GLDM models were employed to capture data complexities and interactions. With an R-squared value of 0.9714 and a Root Mean Squared Error (RMSE) of 17.2027, the fifth-order GLDM model was out to be the most accurate. This model effectively handles non-linear relationships and seasonal trends. Furthermore, the GLDM model's performance was contrasted with that of conventional univariate time series models, including the Prophet, Hybrid AutoARIMA, BATS (Box–Cox transformation, ARMA errors, Trend and Seasonal components), TBATS (Trigonometric seasonality, Box–Cox transformation, ARMA errors, Trend and Seasonal components), Exponential Smoothing, and Multi-Layer Perceptron. The GLDM model outperformed these models in predictive accuracy. The results demonstrate the GLDM approach's resilience and dependability in time series modeling, highlighting its potential to improve forecasting accuracy in the dairy sector. The study concludes that GLDM model significantly outperformance upon existing models, establishing it as a valuable tool for accurate and reliable milk production forecasting.