Lung cancer is one of the most common and lethal cancers all over the world. The identification of the disease at an early stage is important to improve the well-being of patients. Nevertheless, using supervised learning for the problems relating to lung cancer detection has the challenge of an imbalanced dataset as there are costs relating to the misclassification of the different cases. This study looks at the effects of this challenge and balancing techniques in predictive performance enhancement. This research’s primary aim is thus to investigate how effective supervised learning models are in lung cancer detection when applied to both imbalanced and balanced datasets. The application of several data balancing methods, like “SMOTE,” serves the purpose of examining how these changes influence model accuracy, precision, and other performance figures of merit. The methodology involved analyzing the imbalanced dataset to identify the various class distributions, and then subsequent balancing techniques such as oversampling and undersampling techniques were applied. Several supervised learning techniques, such as AdaBoost, Random Forest, Logistic Regression, SVM, Random forest, XGBoost, KNN, and Decision Tree, were performed on both dataset versions while recording performance metrics such as accuracy, F1-score, and ROC. The findings from the balanced dataset exhibited excellent accuracy and precision than the findings of the imbalanced dataset. This research demonstrates the significant benefits of employing dataset-balancing techniques in supervised models geared toward lung cancer detection.