K. Usha Rani, V. Ramakrishna, S.V. Sudheer Kumar and P. Bhargavi
Abstract
Agriculture which forms a major sector of the world economy is increasingly becoming problematic due to climate change, varied soils, and a lack of access to modern technologies. To assist farmers in making better decisions, this paper proposes a machine learning (ML)-based crop recommendation system. The predictive models are trained using past data on cropping, soil parameters (pH, nitrate, potassium and nitrogen), and weather parameters (temperature, precipitation and humidity). The algorithms considered in the evaluation are Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), Bagging (BG), Ada Boost (AB), Gradient Boosting (GB), Extra Trees (ET), and Logistic Regression (LR). Information cleaning techniques such as normalization and noise removal are applied to improve the model. To ensure that models are more robust and can be applied in a broader variety of situations, cross-validation methods are used to further refine the models. Depending on the environmental and agronomic conditions, the system suggests crop recommendations. The results obtained compare ensemble-based methods, and the random forest in particular, to be better than other models. This application of machine learning (ML) enhances crop planning, productivity, and supports the idea of environmentally friendly farming.