AgriFact framework for modelling the impact of farmers’ information demand on nationwide wheat productivity in India

Understanding the link between farmers' information needs and crop yield is vital for crafting effective, sustainable agricultural policies. However, existing research has yet to comprehensively investigate the impact of farmers' information demand on crop yield using advanced analytical tools. In this direction, the presented study introduces the AgriFact framework to explore the relationship between Indian farmers' information inquiries and crop yield using Deep Learning (DL)-based modelling and numerical methods-based variables' relationship analysis. The study examines 1.8 million farmer query calls collected over a decade from Kisan Call Centers, alongside district-wise wheat yield data across India. In the first phase, six DL models are developed and compared to estimate crop productivity based on topic-wise query calls per hectare. From the experiments, it is noted that the 1-D CNN model delivered the highest predictive accuracy, achieving the lowest RMSE (0.759 t/ha) and MAE (0.585 t/ha) among all evaluated models. Later, the study integrates ceteris paribus analysis and factor-wise partial derivatives, demonstrated through a nationwide wheat yield case study. The presented research offers deeper insights into the association between farmers' information demand and wheat crop productivity, potentially informing the formulation of evidence-based agricultural interventions.