A Hybrid CNN-GRU Deep Learning Approach for Accurate Farmer Risk Prediction

Farmers worldwide face numerous hazards that threaten agricultural sustainability, including adverse weather conditions, globalization, pest infestations, and market fluctuations, all of which affect productivity and profitability.Traditional risk prediction methods often fail to adapt to the complexities of modern agriculture.This study proposes an innovative hybrid deep learning (DL) framework that integrates convolutional neural networks (CNN) and gated recurrent units (GRU) in a novel manner to improve farmer risk prediction.Unlike previous approaches that utilize CNN or GRU separately, the proposed model leverages CNN for spatial feature extraction and GRU for capturing temporal dependencies, enabling it to handle dynamic agricultural data effectively.The model is trained on a comprehensive farmer dataset, collected through a nationwide survey that captures a wide array of factors influencing agricultural risk, including climatic conditions, soil health, crop types, farming techniques, and socioeconomic indicators.With 12 distinct features, this dataset provides a robust foundation for predicting and mitigating risks in agriculture.To optimize the feature selection and improve prediction accuracy, L1-norm regularization is applied, ensuring that the model identifies the most relevant factors in real-time risk prediction.The results show that the hybrid model achieves an accuracy of 95.48%, with recall and precision rates of 95.93% and 95.12%, respectively, highlighting its effectiveness.This innovative approach offers significant potential for enhancing agricultural decision-making, helping farmers reduce risks and optimize resource management more effectively.