Sheath blight (ShB), caused by Rhizoctonia solani, is a major constraint in rice (Oryza sativa L.) production, leading to considerable yield losses. The objective of the present study was to assess sheath blight resistance in rice landraces based on the disease progression parameters and machine learning methods. Field screening was conducted under inoculated conditions and disease progression was monitored from 10–70 days after inoculation using disease severity (DS), percent disease index (PDI) and area under disease progress curve (AUDPC). Machine learning models - linear regression, decision tree, random forest and gradient boosting were employed to predict DS of sheath blight in rice landraces based on disease progression parameters and their predictive performances were compared using R², root mean square error (RMSE) and classification-based evaluation metrics. Among the landraces screened, NLR 33892 and Swarna Swoubhagya showed resistance against R. solani, while the majority were susceptible or highly susceptible. Both DS and PDI increased progressively over time and showed a strong positive correlation across assessment stages, confirming consistent disease development. Linear regression performed the best with the highest coefficient of determination (R2 = 0.8178) and the lowest RMSE = 3.852, followed by decision tree (R2 = 0.8077), random forest (R2 = 0.8010) and gradient boosting (R2 = 0.7899). The analysis of confusion matrices further supported the better consistency of classification of linear regression.
