Machine Learning Techniques for Soybean Charcoal Rot Disease Prediction.

Area of Science:

• Agricultural Science
• Plant Pathology
• Computational Biology

Background:

• Charcoal rot disease, caused by *Macrophomina phaseolina*, significantly reduces soybean yield.
• Traditional methods for predicting charcoal rot disease in soybeans are time-consuming and impractical.
• Early pathogen detection is vital for managing disease spread in crops.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for the early prediction of charcoal rot disease in soybeans.
• To assess the effectiveness of a hybrid feature set including physiological and morphological data for disease prediction.
• To identify the best-performing ML model for real-world application in soybean disease management.

Main Methods:

• Development and testing of several ML techniques on a dataset of 2,000 soybean plants (healthy and infected).
• Utilized a combined set of physiological and morphological plant features as input for the ML models.
• Evaluated model performance using metrics such as accuracy, sensitivity, and specificity.

Main Results:

• All developed ML models achieved over 90% accuracy in predicting charcoal rot disease.
• Gradient Tree Boosting (GBT) demonstrated superior performance with 96.25% sensitivity and 97.33% specificity.
• The inclusion of physiological features significantly improved the predictive power of the ML models.

Conclusions:

• Machine learning, especially Gradient Tree Boosting, is highly applicable for predicting soybean charcoal rot disease in practical settings.
• Physiological features are important indicators for enhancing the accuracy of ML-based plant disease prediction.
• The study provides a robust ML framework and dataset for future research in soybean disease management.