Abstract

Agriculture is extremely vital to India's economy. Farmers' main concern is that they do not choose the proper crop based on factors such as soil nutrients, humidity, water level, moisture, and season. As a result, they are experiencing a major drop in productivity. Machine learning algorithms are employed in modern farming operations to select the best crops based on soil types, weather, and climatic conditions. In this study, a model is developed to forecast feasible crops based on the soil physicochemical properties, climate and crop rotation factors. An optimized XGBoost classifier model is developed using the Whale Optimization (WO) algorithm which would choose appropriate hyperparameters for the boosting tree classifier. Further, feature selection algorithm is employed which would improves the accuracy by selecting optimal feature subset by eliminating insignificant and redundant features. The crop recommendation using WO based XGBoost achieves the accuracy of 95.7 % which is significantly higher than the XGBoost and Grid Search-XGBoost.

Authors

P. Nithya¹, A.M. Kalpana², P. Tharani³
Government College of Engineering, Salem, India^1,3, Government College of Engineering, Dharmapuri, India²

Keywords

Machine Learning, Classifier, Optimization, Feature Selection, Crop Recommendation