Estimativa do valor justo para agricultores no agronegócio avícola usando modelos de machine learning

Abstract

This article aimed to estimate the fair value that farmers should re ceive in transactions within the poultry agribusiness, using machine learning models. The research covered topics such as predictive analysis, machine le arning, poultry farming, and fair value accounting. The methodology inclu ded data collection through interviews and invoice analysis, complemented by a synthetic dataset. Initial tests with models such as Random Forest and Gradient Boosting, among other models, resulted in low R² values. However, adjustments considering the capacity of the poultry houses improved the predictive perfor mance, achieving R² values ranging from 0.86 to 0.90. The analysis of variance highlighted the relevance of certain variables. Despite the limitations of synthe tic data, the results indicated the feasibility of predicting fair value using ma chine learning. Future validation with real datasets is recommended for greater accuracy.