Assessing Performance of the Transformer Model in Predicting Hog Prices

Abstract

The price of hogs has a significant impact on livelihoods, social development, and overall stability. Therefore, accurate prediction of hog prices is crucial for effective decision-making, breeding strategies, resource allocation, and risk mitigation. In this study, we compare the performance of Transformer and Recurrent Neural Network (RNN) models in predicting hog prices and evaluate their applicability in different scenarios. Additionally, we conduct a generalization test on the hog pig industry chain to assess the models' performance. Our findings indicate that Transformer models excel in parallel computing, context capture, and encoding/decoding tasks. On the other hand, RNN models demonstrate superior performance in predicting extreme events and localized tasks. Therefore, the choice of modeling method should be tailored to meet specific requirements based on the nature of the prediction task.