Lung Cancer Classification with XGBoost, SVM, Random Forest, and k-Nearest Neighbor

Author : Dr.Mitat Uysal , Dr.Aynur Uysal

Abstract : Early identification of lung cancer risk from tabular clinical measurements can support decision-making by prioritizing patients for imaging and further diagnostic workflows. This paper presents a comparative study of four widely used classifiers—k-Nearest Neighbor (kNN), Support Vector Machines (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost)—for lung-cancer classification. We summarize the mathematical foundations of each method, discuss practical considerations such as feature scaling, class imbalance, and overfitting, and provide a reproducible Python pipeline implemented without sklearn, tensorflow, or networkx. For data, we focus on the classic UCI Lung Cancer dataset (32 instances, 56 features) as a compact benchmark for algorithmic comparison, while emphasizing the limitations of small-sample medical datasets and the need for external validation.

Keywords : Lung cancer, classification, kNN, SVM, Random Forest, XGBoost, gradient boosting, CART, ROC-AUC, confusion matrix, medical ML.

Conference Name : International Conference on Machine Learning and Artificial Intelligence (ICMLAI - 26)

Conference Place : Vienna, Austria

Conference Date : 19th Feb 2026