Hybrid Deep Learning Model for Crop Yield Prediction Using CNN and Random Forest

Abstract

Crop yield prediction is fundamentally related to food security and sustainable agriculture. This work proposes a hybrid deep learning model, integrating CNN and RF for improving the crop yield prediction accuracy. The CNN model extracts complex spatial and temporal features from agricultural datasets, while the Random Forest model performs robust ensemble-based prediction. The system is trained using climatic parameters comprising rainfall, temperature, humidity, and soil nutrient content (NPK). A comparative analysis with traditional models comprising Linear Regression, Decision Tree, and LSTM indicates that the proposed CNN–RF hybrid model achieves superior performance with an accuracy of 94.6%. The results indicate the efficacy of the proposed approach for precision agriculture and help drive data-driven decisions for farmers and policymakers.