Classification of soybean genotypes during the seedling stage in controlled drought and salt stress environments using the decision tree algorithm

Alan Mario Zuffo,Fábio Steiner,Jorge González Aguilera,Heber Hernán Soto Gonzales,Bruno Rodrigues de Oliveira

DOI: https://doi.org/10.1111/jac.12654

2023-06-02

Journal of Agronomy and Crop Science

Abstract:Soybean is one of the most important oilseed crops grown worldwide. However, abiotic stresses such as drought and salinity can seriously affect soybean production, especially in tropical climate conditions. To evaluate the adaptability and stability of soybean genotypes under abiotic stress conditions, some studies have proposed a multitrait tool to select stress‐tolerant soybean genotypes through a multitrait stability index (MTSI). This index can be used under stressful environmental conditions to quantify the genotypic stability of soybean cultivars. Our study is based on an unprecedented approach, where we propose to use a machine learning algorithm called 'Random Forest' to obtain a classification model based on a decision tree algorithm. The decision tree data structure can be used even by nonexperts facilitating the decision‐making process for genotype selection. The proposed model evaluated the importance of six shoot and root morphological variables and predicted from which controlled growth environment the soybean plants originated. Using this model more than 73% of the genotypic patterns were learned correctly. Besides that, this model can also predict and rank the most critical variables in the development of soybean genotypes, having obtained results very similar to recent field research. The research is important for plant breeders who seek an early selection of soybean seedlings for drought and saline stresses.

agronomy