Enhancing Lung Cancer Early Detection: A Hybrid Ensemble Model

Abstract

Cancer stands as one of the leading causes of death diseases worldwide; it, therefore, must be complex and multi-faceted. Basically, cancer is characterized by uncontrolled cell growth with a tendency to invade or spread to surrounding tissues. Among such cancers, lung cancer stands as one of the most prevalent and deadliest. Accordingly, the identification of risk factors and early symptoms can greatly improve survival rates and quality of life for patients. In general, individual choices of lifestyle—smoking habits, dietary intake, level of physical activity, and exposure to environmental toxins—are very significant in relation to the risk of developing lung cancer. Their evaluation cannot be underemphasized while assessing prevention and early detection strategies. Data like this on lifestyle and symptoms can become very important for researchers in coming out with the trends and correlations of lung cancer development. The application of machine learning has demonstrated significant potential in improving diagnostic precision in the medical domain, specifically in the identification and categorization of diverse cancer types. Improving patient outcomes and survival rates requires early and precise detection. One strategy that can be used in lung cancer detection for the purpose of improving precision, this study investigates the creation of an ensemble model including Decision Tree (DT), Random Forest (RF), and Artificial Neural Networks (ANN) classifiers. By summing up these prediction powers of every individual model, the ensemble technique makes use of their strengths to produce an effective diagnostic tool. To train and validate the model, a variety of patient data, including symptoms and lifestyle characteristics, were included in the dataset used in this work. The results indicate that the ensemble model performs better compared to individual classifiers. This is shown during model evaluation, where the Ensemble model could detect the absence or presence of lung cancer with an accuracy of 98.37%.