AgriCrop - Health Monitoring and Recommendation of Crop Based on Soil and Weather Parameter

The traditional ways of farming have been on manual observation and intuition of the farmer who is frightened to detect the stress of the crops, lack of balance of nutrients and improper selection of the crops. These restrictions lead to low productivity, economic wastages, and poor utilization of farming resources. The proposed paper is the data-driven, integrated machine learning framework that runs on Decision Tree, XGBoost (Extreme Gradient Boosting) to accurately predict crops and assess their health through the help of soil and environmental characteristics. In contrast to the earlier methods, which placed a major emphasis on the simpler ensemble algorithms like Random Forest, this paper goes to show that XGBoost is far much ahead of the rest of the algorithms since it uses sequential tree boosting, regularization, and a highly optimized calculation of the gradient, yielding a prediction Accuracy is 99.03 %, Recall is 94.1 %, Precision is 94.2 % and the $\mathbf{F 1}$-Score is 94.2 %. In comparison of the runner up model i.e. Random Forest Classifier with Accuracy 93.5%, Recall 88.5%, Precision 82% and F1-Score is 85.1%. The model predictions are transformed into practical guidance such as fertilizer recommendations, $\mathbf{p H}$ corrections, irrigation schedules and nutrient balancing by an inference engine that operates utilizing the principles. Experimentally, this has been confirmed since XGBoost was the most precise, generalization model, and error-correcting model which is the most stable model to be employed in precision agriculture in non-IoT setup. The proposed system is expandable, understandable and capable of accommodating the area of application in the diverse farming environments.