Lung cancer prediction model using ensemble learning techniques and a systematic review analysis

Abstract

Lung cancers are malignant lung tumors resulting from uncontrolled growth of lung cells that metastasizes to other parts of the body and can cause death. Although lung cancer cannot be prevented, the risk of cancer development can be lowered. Early detection of lung cancer is essential for patient survival, and machine learning-based prediction models have potential use in predicting lung cancer. Ensemble techniques are compelling and powerful techniques in Machine Learning to improve the prediction accuracy as classifiers. This paper reviewed some research articles on lung cancer prediction models that used machine learning and ensemble learning techniques. Furthermore, we added our newly developed ensemble learning techniques to this paper which was developed based on a survey dataset of 309 people with or without lung cancer by oversampling SMOTE method. The ensemble techniques we used are XGBoost, LightGBM, Bagging, and AdaBoost by k-fold 10 cross-validation method and the attributes our lung cancer prediction models used are age, smoking, yellow fingers, anxiety, peer pressure, chronic disease, fatigue, allergy, wheezing, alcohol, coughing, shortness of breath, swallowing difficulty, and chest pain. Results: According to our analysis, the XGBoost technique performed better than other ensemble techniques and achieved an accuracy of 94.42 %, precision of 95.66%, recall of 94.46%, and AUC of 98.14%, respectively.