SOIL CLASSIFICATION USING MACHINE LEARNING FOR CROP SUGGESTION

Abstract

Abstract: Soil plays a prominent role and the appropriate soil selection for a crop is a significant aspect of agriculture. Progression of agriculture and yield can be increased with the assistance of computational algorithms. Machine Learning (ML) provides a pivotal tool for Decision Support System (DSS) in soil classification and recommendation systems. Numerous, ML algorithms have been deployed in the classification of soil and recommendation of crops. In this research work, an efficient Gradient Boosted Tree (GBT) is utilized in feature selection and classification is accomplished using Feed Forward Neural Network (FFNN). This work considers ten different soils and an appropriate crop is recommended for soil. The framework of GBT is efficient and robust in handling data with nonlinear relationships. The FFNN’s scalability, adaptability, and interpretability offer the highest accuracy while classifying complex data. Finally, the Soil-Crop suitable Matrix (SCM) is constructed to map the suitable crop source for appropriate soil, whereas the scores are generated for distinct crops for a specific type of soil. The performance of the proposed Machine Learning assisted Soil Classification and Crop Recommendation Framework (ML-SCCRF) is compared with existing state-of-art techniques, where the ML-SCCRF outperforms the existing frameworks.