<p>Crop recommendation systems use factors like soil type, climatic conditions, and historical data in recommending suitable crops for a particular place with the help of various machine learning and deep learning models. The paper proposes a novel crop recommendation system using a Deep Convolutional Neural Network (DCNN), in which the weight updation of DCNN is done using an enhanced Honey Badger Optimization Algorithm (LOLSHBA). The conventional HBA may suffer from some drawbacks, including slow convergence speed, unbalanced exploration and exploitation, and stagnation on local optima. Thus, for training the DCNN to perform accurate crop predictions and efficient recommendations, we use an enhanced variant of HBA, namely LOLSHBA. Numerical input features such as nutrients in the soil (N, P, K), pH, temperature, humidity, rainfall, and precipitation are drawn from three publicly available Kaggle datasets. Extensive preprocessing of these datasets has been performed to handle missing values, eliminate duplicate records, and maintain structural consistency, thereby enhancing data quality before model training. Further, crop recommendation is performed using the DCNN–LOLSHBA framework. Experimental results demonstrate that DCNN–LOLSHBA consistently outperforms existing algorithms in terms of accuracy, precision, recall, F1-score, Matthews correlation coefficient (MCC), and execution time across three different crop recommendation datasets. Specifically, the proposed model achieves up to 99.68% accuracy, 99.66% precision, 99.78% recall, 99.48% F1-score, and 98.63% MCC, thereby confirming its effectiveness and robustness. By tapping into data-driven insights, this system provides reliable and personalized recommendations for selecting suitable crops under specific environmental and soil conditions.</p>