COMPARATIVE ANALYSIS OF MACHINE LEARNING TECHNIQUES FOR ACCURATE LUNG CANCER PREDICTION

Аннотация

INTRODUCTION robustness in predictive tasks due to their ability Lung cancer remains one of the most prevalent and to handle large-scale data and complex deadly forms of cancer globally, accounting for a interactions between features (Chen, Song, & significant number of cancer-related deaths each Zhang, 2020; Ke et al., 2017). year (Siegel, Miller, & Jemal, 2023). The challenge of early detection and accurate prediction of lung The utility of these algorithms in cancer prediction cancer has driven extensive research into the use is underscored by recent research highlighting of machine learning (ML) algorithms to enhance their effectiveness in various contexts. For diagnostic capabilities and improve patient example, Khan et al. (2023) have illustrated the outcomes. The advent of advanced computational potential of XGBoost and LightGBM in breast techniques has opened new avenues for analyzing cancer detection, providing a basis for their complex medical data, leading to significant application in other cancer types, including lung progress in cancer prognosis and classification. cancer. Similarly, other studies have evaluated the performance of different classifiers in predicting Machine learning offers a promising approach to myocardial infarction, underscoring the predicting lung cancer by leveraging large datasets importance of choosing the right model for specific and sophisticated algorithms to uncover patterns medical conditions (Khan, Miah, Abed Nipun, & that might not be immediately apparent through Islam, 2023). traditional methods. Recent advancements in ML, particularly in algorithms such as XGBoost, Despite the promising results of existing studies, LightGBM, AdaBoost, Logistic Regression, and the application of ML algorithms to lung cancer Support Vector Machines (SVM), have prediction remains an evolving field. The demonstrated their potential in various medical complexity of lung cancer data, which includes a applications. For instance, studies have shown that range of clinical, demographic, and environmental XGBoost and LightGBM, both gradient boosting factors, necessitates a thorough evaluation of frameworks, provide high accuracy and different algorithms to determine the most https://www.theamericanjournals.com/index.php/tajet 93 THE USA JOURNALS THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN – 2689-0984) VOLUME 06 ISSUE09 effective approach for accurate prediction. Recent myocardial infarction prediction, illustrating the work has highlighted the importance of not only challenges and opportunities in predictive achieving high accuracy but also considering modeling for health outcomes (Khan, Miah, Abed metrics such as sensitivity, specificity, and F-1 Nipun, & Islam, 2023). This work emphasizes the score to ensure comprehensive model evaluation importance of evaluating various classifiers to (Xia et al., 2023). determine the best fit for specific medical conditions, a concept that is critical for lung cancer In this study, we aim to build upon the existing prediction as well. body of research by providing a detailed comparison of several ML algorithms in the Recent advancements have highlighted the efficacy context of lung cancer prediction. By evaluating the of gradient boosting algorithms in cancer performance of XGBoost, LightGBM, AdaBoost, prediction. For example, Chen et al. (2020) Logistic Regression, and SVM based on accuracy, investigated the performance of XGBoost in sensitivity, specificity, and F-1 score, we seek to predicting cancer outcomes, demonstrating its identify the most effective tools for clinical superior capability in handling complex datasets application. Our approach includes an in-depth and providing accurate predictions (Chen, Song, & analysis of attribute correlations and model Zhang, 2020). Similarly, LightGBM has been noted performance, contributing to a more nuanced for its scalability and efficiency, especially in large- understanding of each algorithm's strengths and scale datasets, which is crucial for handling diverse limitations in predicting lung cancer. patient data (Ke et al., 2017). LITERATURE REVIEW In contrast to these studies, our research distinguishes itself by focusing specifically on the The application of machine learning (ML) application of multiple ML algorithms to lung techniques to improve lung cancer prognosis has cancer prognosis, with an emphasis on evaluating been an area of extensive research, with several not only accuracy but also sensitivity, specificity, studies exploring different algorithms and and F-1 score. While previous studies have methodologies. explored various algorithms and their general Early research in this domain has demonstrated applications, our work provides a comprehensive the potential of ML algorithms in cancer detection comparison of XGBoost, LightGBM, AdaBoost, and classification. For instance, Khan et al. (2023) Logistic Regression, and Support Vector Machines explored various ML algorithms for b