Smart agriculture leverages technologies such as the Internet of Things (IoT), artificial intelligence (AI), and big data analytics to enhance sustainability, productivity, and resource efficiency. This study proposes a novel Permutation Flamingo Optimized Recommendation System (PFO-RS) that integrates deep learning with explainable AI principles to improve transparency and decision-making in precision farming. The proposed framework employs the Flamingo Optimization Algorithm for permutation-based feature selection across complex agricultural datasets comprising soil nutrients, crop type, and climatic variables. These optimized features are then used by a recurrent neural network with long short-term memory (RNN-LSTM) to generate adaptive recommendations on optimal crop selection, irrigation scheduling, and fertilizer application. Experimental evaluation using multi-year agricultural datasets from South India demonstrated that PFO-RS achieved superior performance compared with baseline models such as SVM, Decision Tree, XGBoost, and CNN, with an average accuracy improvement of 2.5–3.0% across five major crops and an R² value of 0.98 for yield prediction. Mean absolute error (MAE) and root mean square error (RMSE) were reduced to 0.092 and 0.089, respectively, compared to 0.108 and 0.126 for the best-performing conventional models. Field-level validation indicated a 5–8% improvement in predicted productivity when applying the model’s recommendations in simulated real-world farming conditions. These results confirm that the proposed PFO-RS framework provides accurate, explainable, and scalable support for data-driven agricultural decision-making.