AI-Powered Smart Agriculture Decision Simulation and Prediction System

Abstract

Artificial intelligence and machine learning techniques have become essential for developing intelligent agricultural systems capable of improving crop selection, yield estimation, and climate-based decision-making. Traditional agricultural practices often rely on manual analysis and historical experience, which may not effectively handle complex relationships among soil parameters, environmental conditions, crop characteristics, and rainfall variations. This limitation motivates the development of an AI-powered smart agriculture decision simulation and prediction system for accurate agricultural forecasting. The system utilizes three datasets obtained from Kaggle, including the Crop Recommendation Dataset, Rainfall in India dataset covering the period from 1901 to 2015, and Crop Yield in Indian States dataset. The datasets are processed through null value removal, duplicate elimination, feature selection, label encoding, format transformation, and time-series feature generation. Multiple machine learning and deep learning models are implemented, including Random Forest, Multilayer Perceptron, Gradient Boosting, KNN, Logistic Regression, CNN, XGBoost, CatBoost, AdaBoost, Lasso, Ridge, Stacking Regressor, Voting Classifier, and ARIMA-based rainfall forecasting. Model performance is evaluated using accuracy, precision, recall, F1-score, MSE, RMSE, MAE, and R² metrics. The best-performing models are identified through comparative analysis and enhanced using Explainable AI techniques such as LIME and SHAP for interpretability. The developed framework provides reliable, transparent, and data-driven agricultural predictions for improved farming decisions.