Optimizing Crop Recommendations: A Novel IDCDMO-Enhanced ACGRU Approach for Advanced Agricultural Predictions

Abstract

Agricultural productivity is maximised to a greater extent by crop recommendation. The traditional methods of crop recommendation are often rely on limited data, simple models, heuristic rules and expert knowledge which are less dynamic and not fully address the agricultural challenges.  Deep Neural Networks (DNN) play an important role in crop recommendations by providing more accurate crops that are suited to the terrain. The research proposes a novel approach utilizing Attention based Convolutional Gated Recurrent Unit (ACGRU), neural network architecture for crop recommendation. Additionally it integrates Improved Distribution based Chaotic Dwarf Mongoose Optimization (IDCDMO) algorithm for feature selection. This study compares, for crop recommendation, to see the effectiveness of proposed approach. The performance of the IDCDMO¬¬-enhanced ACGRU is compared with ACGRU without feature selection and also with the conventional neural network models such as Feed Forward Neural Network (FNN) and Long Short Term Memory (LSTM). The IDCDMO-enhanced ACGRU significantly outperforming both the ACGRU without feature selection and conventional neural network models in terms of accuracy, precision, recall and f1 score. Therefore, the integration of IDCDMO with ACGRU efficiently enhances recommendation accuracy, improves soil health and adds greater agricultural productivity.