Drought and Salinity Stress Classification in Soyabean Crops: Comparative Analysis of Machine Learning Models

Soyabean cultivation encounters significant obstacles attributed to abiotic stresses like drought and salinity, especially in areas characterized by tropical climates. In this paper, we seek to fill gaps identified in previous research efforts, particularly the lack of comprehensive analysis regarding genotypic factors and the underutilization of morphological characteristics in machine learning algorithms for Soyabean genotype selection. We propose a novel approach to bridge these gaps by employing six morphological variables in machine learning models. Through the process of training the models using specific parameters, plants are categorized into either the Control group or the Stress Group, which comprises of genotypes facing drought or salinity stress. Various models such as Random Forest Classifier (RFC), Support Vector Machine (SVM), Gradient Boosting (GB), k-Nearest Neighbors (kNN) and Naive Bayes algorithms are included, offering an in-depth comparative analysis of the performance and feature importance under each model used. However, the Support Vector Machine was identified as the most accurate algorithm with an accuracy of 96.79% in stress classification of Soyabean genotypes. Incorporating morphological parameters in machine learning models enhances the accuracy and efficiency of Stress Classification process, increasing the sustainability of Soyabean production and its ability to withstand environmental adversities.