Effective Regional Prediction for Potato Stress Conditions Across Rabi Seasons

West Bengal is a major potato belt in India where the crop is grown typically in the Rabi season. Stress conditions such as Late Blight, Stem Rot, and Thrips significantly impact crop output. Cultivating the crop in a given region follows certain seasonal patterns. However, variations in components like environment, climate-change effects, evolving crop history, and management practices are observed across seasons. Forecasting models developed for a given season in a region need to exhibit robustness for subsequent seasons by quickly learning relatively stationary factors and adapting quickly to dynamic ones. We propose and evaluate prominent forecasting models and strategies across the Rabi seasons of 2020–21 and 2021–22 in West Bengal. The emphasis here is to assess model performance for the subsequent season after it has seen the previous one. We propose a model to forecast the incident pattern envelope using a supervised learning component for which ANN, XGBoost, and Random Forest (RF) are evaluated. RF-based model is found to perform better when compared to the other models, and it could be useful for taking macro-level decisions related to agricultural input planning on the ground. This approach is then used to generate results for other stress conditions to assess overall benefits.