COMPUTATIONAL INTELLIGENCE FOR THE CLASSIFICATION OF ADAPTABILITY AND STABILITY IN COMMON BEAN GENOTYPES

Abstract

Genotype-by-environment (G×E) interactions complicate the recommendation of common bean (Phaseolus vulgaris L.) varieties. This study aimed to explore the potential of the Extreme Gradient Boosting (XGBoost) machine learning algorithm as a complementary tool for classifying the adaptability and stability of these genotypes in breeding programs. The methodology involved training the XGBoost model with simulated data aligned with the classes proposed by Eberhart and Russell (1966) and Finlay and Wilkinson (1963), followed by the classification of real genotypes. Eleven genotypes were evaluated across 32 environments (combinations of location and year) using a randomized block design. XGBoost showed agreement rates of 100% and 81.82% with the Eberhart and Russell (1966) method for adaptability and stability classifications, respectively. Additionally, the algorithm achieved an accuracy of 89.3%. These results demonstrate the machine learning model's capability to classify genotypes according to established patterns, offering greater analytical flexibility. However, it is crucial to consider that the model's practical applicability and performance are directly linked to the quality and representativeness of the training data. In summary, XGBoost emerges as a promising tool for optimizing adaptability and stability analyses, enhancing breeding efficiency, and improving the precision of cultivar selection.