Deep Belief Network Optimization Using PSOGAA Algorithm for Efficient Crop Recommendation

The agricultural sector plays a crucial role in the development of an emerging economy, providing employment to a significant portion of the population. Selecting the most suitable crops based on specific weather and soil conditions is vital to maximize yield and ensure profitability for farmers, contributing to global food security. In this context, this paper introduces a crop recommendation model utilizing Deep Belief Networks (DBN), fine-tuned by a hybrid optimization approach combining Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Adam optimization. To enhance the efficiency of DBN-based crop recommendation model, this paper proposes the PSOGAA algorithm, a novel hybrid optimization algorithm designed to optimize the DBN’s network structure. Subsequently, genetic operators featuring self-adjusting crossover and mutation probabilities are applied to further refine the PSO and explore a global optimization solution. Ultimately, this global optimization solution is employed to construct the network structure of the crop recommendation model. Experimental results demonstrate the superior performance of the proposed algorithm compared to other DBN optimization techniques, showcasing an average classification accuracy improvement of at least 2.3%. This emphasizes the effectiveness of the PSOGAA algorithm as an efficient DBN optimization technique. Focusing on four major crops in India—rice, sugarcane, maize, and finger millet—the crop recommendation model’s outcomes are compared with those of the DBN model optimized using alternative algorithms. The experimental findings highlight that the proposed hybrid optimization algorithm surpasses other optimizers, exhibiting enhanced accuracy in crop recommendations.