Predictive Modeling of Paddy Crop Yields in Jhapa, Nepal: A Machine Learning Approach for Smallholder Agricultural Households.

Agriculture forms the backbone of Nepal’s economy, yet smallholder farmers in the rural region face increasing uncertainty in crop production due to climate variability and limited access to data-driven decision tools. This study investigates the applicability of lightweight machine learning models for predicting paddy yields in Jhapa district, Nepal. By integrating primary survey data collected from 30 smallholder farmers with secondary meteorological records from 2019 to 2023, the research evaluates the performance of Multiple Linear Regression and Decision Tree regression models. The results indicate that irrigation access and rainfall distribution are the most influential predictors of yield, with the Decision Tree model achieving higher predictive accuracy (R² = 0.87). The findings demonstrate that interpretable, low-cost AI models can provide meaningful insights even in data-scarce environments, offering a scalable framework for agricultural decision-support systems in developing regions. This record includes an anonymized CSV dataset of 30 farmers to ensure research transparency while protecting participant privacy.