Optimized Resampling Techniques for Crop Recommendation Systems in Mizoram Agriculture

This study employs the Logistic Regression (LR) model to examine various resampling techniques for determining the optimal methods that tackle the data imbalance in the dataset used. Based on the observation result from various metrics, the Support Vector Machine Synthetic Minority Oversampling Technique (SVMSMOTE) achieved the best results (1 in the test, 0.998 in CV scores across the metrics) and was used for training the models. Then, Adaptive Boosting Classifier (AdaBoostC), Light Gradient Boosting Machine Classifier (LightGBMC), and Support Vector Machine Classifier (SVMC) were trained on SVMSMOTE resampled data. Further, K-fold cross-validation (with K =5) was employed to assess the generalizations of the models, and Optuna for tuning hyperparameters. Eventually, the performance of the models was evaluated using various metrics such as precision, recall, f1-score, and accuracy. Based on the experimental results, the SVMC consistently outperformed the other models, achieving 1 (i.e., perfect performance) on the test and 0.998 on the CV scores in all metrics.