Multi-Criteria AI System for Crop and Fertilizer Prediction
Abstract
The significance of this research is the improvement of the forecasting precision for crop and fertilizer, and it specifically compares the performance difference between the Random Forest and LightGBM models. The study is conducted based on a Kaggle agricultural dataset comprising 10,000 records, where much information is given on the nature of the soil, climate, and crops. The dataset was preprocessed to handle missing values, normalized, and divided into training and testing sets. Random Forest and LightGBM models were trained using the same input features, and their performances were evaluated based on accuracy, precision, recall, and F1-score. The LightGBM model outperformed the Random Forest model, achieving an accuracy of 89.48%, precision of 67%, recall of 61.28%, and F1-score of 62%, compared to 85.02%, 64%, 33.13%, and 60% for Random Forest, respectively. The difference was statistically significant (p < 0.001). The results indicate that LightGBM provides higher prediction reliability and stability, making it more suitable than Random Forest for smart agriculture systems and decision-making regarding crop selection and fertilizer recommendation.