Revolutionizing Lung Cancer Detection: A High-Accuracy Machine Learning Framework for Early Diagnosis

Lung cancer is a deadly disease. According to a report of 2024, it is the primary reason for 1.82 million deaths. Given the high disease burden, early detection of lung cancer is crucial for improving survival rates and implementing effective strategies. This paper is aimed at conducting a systematic literature review and developing a highly accurate framework for predicting lung cancer effectively. Tollgate methodology has been used for systematic literature review, and quality assessment criteria were applied to select published articles relevant to the research questions. The paper investigates the effectiveness of machine learning in identifying patterns relevant to lung cancer prediction (Q1), examines the pros and cons of current predictive systems (Q2), compares the use of artificial intelligence in lung cancer prediction with traditional methods (Q3), and identifies key features that distinguish lung cancer from patient symptoms (Q4). Machine learning techniques were employed for the proposed framework. Two publicly available, distinct datasets containing clinical features were obtained. Then, the SelectKBest method was used for feature selection, and SMOTE was used to handle class imbalance. Our proposed framework includes a voting ensemble with random forest, support vector machine, and logistic regression with cross-validation. The results indicate an accuracy of 99% and 92.5% for the first and second datasets, respectively. This study's systematic literature review, based on four research questions and a machine learning model, exhibits high accuracy in predicting lung cancer.