Accurate Yield Forecast for Sustainable Agriculture

Food security around the world is being threatened by climate change as well as the overuse of pesticides within the agricultural sector. This research provides an automated crop yield prediction system based on machine learning (ML) for the years 1990 to 2023, taking into consideration historical yields, meteorological data, and pesticide usage. The chosen models were XGBoosting, K-Nearest Neighbors (KNN), and Linear Regression owing their successes to complex pattern recognition, speed, and ease of understanding, respectively. GridSearchCV with K-Fold Cross-Validation was utilized to fine-tune hyperparameters to increase precision and generalization. Data scarcity issues were dealt with by generating synthetic data which increased performance in low-data situations. The best algorithm was chosen by evaluating all models based on predictive accuracy, and metrics such as prediction accuracy, overfitting, scalability across crops and regions were met. This system enables more reliable yield predictions, fostering progress towards sustainable agriculture, relevant to the altered climate conditions.