Enhancing Crop Recommendation with Conformal Prediction and Calibrated Confidence
Students & Supervisors
Student Authors
Supervisors
Abstract
The dynamic nature of the global population is increasing, and significant demand for sustainable farming requires the creation of smart decision support-systems in crop recommendation. A comparative summary of eleven supervised machine learning models, that include Extra Trees (ET), Random Forest (RF), Naive Bayes (NB), CatBoost (CB), XGBoost (XGB), LightGBM (LGBM), Gradient Boosting (GB), Support Vector Machine (SVM), Decision Tree (DT), K- Nearest Neighbors (KNN), and Logistic Regression (LR), was done in this research about the multiclass crop prediction. The publicly providing Crop Recommendation Dataset is used in the study that includes soil nutrient (N, P, K), and environmental (temperature, humidity, pH, rainfall) parameters that are considered crucial. Accuracy, precision, recall, and F1-score were used as measures of model performance, and 5-fold cross-validation was used in evaluation of model stability. The Extra Trees classifier achieved the highest accuracy of 99.72%, outperforming other models in predictive power and generalization. Feature importance analysis highlighted soil nutrients and rainfall as dominant factors, with interpretability enhanced through LIME. The proposed framework integrates optional PCA-based dimensionality reduction for compact modeling, ensemble classifiers for high accuracy, and LIME-based explanations in the original feature space to ensure transparent decision support.
Keywords
Publication Details
- Type of Publication:
- Conference Name: International Conference on Computer and Information Technology (ICCIT 2025)
- Date of Conference: 19/12/2025 - 19/12/2025
- Venue: Long Beach Hotel, Cox’s Bazar, Bangladesh
- Organizer: IEEE