A novel hybrid DNN-RNN framework for precise crop yield prediction

Crop yield forecasting is essential for agricultural planning, resource allocation, and decision-making. Due to crop genotype, environmental conditions, management practises, and their interplay, crop yield prediction is difficult. For the purpose of predicting crop yields, we present a novel hybrid Deep Neural Network-Recurrent Neural Network (DNN-RNN) framework. We aim to employ of the beneficial abilities of DNN and RNN concepts in gathering both temporal and spatial information. We first gather crops data and perform a Z-score normalisation. Followed with Independent Component Analysis (ICA) is employed to extract the relevant features and the particle swarm optimization (PSO) approach to choose the most appropriate feature. To determine how well the suggested hybrid framework performs, extensive experiments are conducted using real-world crop yield datasets evaluation criteria MAE and accuracy are used to assess the prediction accuracy of the hybrid structure. The parameter used to evaluate the outcomes of proposed technique in regards of accuracy (97%), Precision (99%), Recall (97%), and F1-score (98%) MAPE (4.57), MSE (0.0333), MAE (0.4173), Sensitivity (99.73%), Specificity (100%).The hybrid approach captures crop yield prediction's complex dynamics by integrating geographical and temporal information. This framework can help farmers as well as others enhance agricultural practices for production.