Development of Efficient Forecasting Models for Climate-Resilient Crop Rotation Based on Deep Learning Techniques

Crop rotation plays a pivotal role in fostering sustainable agriculture, preserving soil health, mitigating pest and disease issues, and enhancing crop productivity. However, the challenges posed by climate change have made it increasingly herculean tasks for farmers to formulate and implement effective crop rotation strategies. Deep learning holds the potential to usher in a transformation in the realm of crop rotation planning. It can enable the creation of highly efficient classifiers capable of forecasting the adaptability of various crop combinations to the changing climate. These classifiers, in turn, can be harnessed to identify the most resilient crop rotation plans tailored to individual farms and specific climate scenarios. The proposed research work offers an in-depth exploration of the convergence between deep learning methods and the prediction of crop rotation in the context of climate-resilient agriculture. The study uses a curated Kaggle dataset comprising approximately 2200 samples, encompassing an array of agricultural, meteorological, and environmental data. The primary focus of this research centers on four distinct deep learning techniques: deep neural networks (DNN), recurrent neural networks (RNNs), 1D convolutional neural networks (CNN1Ds), and gated recurrent units (GRUs) for crop rotation prediction. The performance evaluation was carried out in terms of accuracy, F1-score, precision, and recall derived from the best model. Based on the outcome, the gated recurrent unit (GRU) outperformed the other deep learning models with an accuracy of 98%.