Enhancing urad bean (Vigna mungo L.) crop management with machine learning: Predictive analysis of pod rot severity and pod bug incidence patterns

Abstract

Urad bean (Vigna mungo L.), commonly known as black gram, is an important pulse crop in Indian agriculture. However, the crop confronts significant challenges due to diseases, including pod rot caused by Fusarium sp, and pest attacks by the pod bug (Clavigralla gibbosa). Accurate prediction of disease severity and pest incidence is essential for formulating effective management strategies to ensure sustainable crop production. A comprehensive field experiment was conducted at the Crop Research Center, Pantnagar, Uttarakhand, during the rainy seasons of 2021 and 2022. The primary objective was to analyze the behavioral patterns of disease severity and pod bug infestations in urad bean. Data on pod rot disease severity and pest incidence were meticulously recorded on a weekly basis. Four Machine Learning approaches, namely ANN, Lasso, Ridge, and Random Forest, were trained and tested to understand the influence of meteorological parameters on pod rot and pest severity. The Random Forest model exhibited superior generalization performance in predicting both disease severity and pest incidence, closely followed by Ridge regression and Lasso regression. The ANN model showed slightly higher testing error metrics. Notably, the Random Forest model demonstrated effective control overfitting, yielding maximum R-squared values of 0.70 and 0.82 for pod rot and pest incidence, respectively. The study’s findings offer valuable insights for agricultural stakeholders in selecting appropriate prediction models to optimize crop management practices and promote sustainable agriculture.