Impact of Lung Data Augmentation Through Instance Replication on KNN Classification Performance

Resampling is a commonly employed technique to enhance classifier performance in the presence of imbalanced datasets. However, its efficiency diminishes when applied to small-sized datasets, presenting challenges in evaluation. Resampling, being adaptable to any base learner, offers a versatile solution for improving classification accuracy. This study investigates the impact of resampling on KNN classifiers across various split percentages using a lung cancer dataset with varying levels of imbalance. The focus is on evaluating the performance of the K-Nearest Neighbors (KNN) classifier under different replication levels. Performance metrics such as accuracy, precision, recall, F1-score, and ROC score are calculated for each replication percentage. The results demonstrate that resampling consistently enhances the classifier's performance as replication levels increase. Detailed interpretations of these metrics are discussed, emphasizing the efficacy of resampling in improving predictive accuracy for lung cancer classification. This study highlights the potential of resampling as a reliable tool for addressing data imbalance and optimizing classifier outcomes.