Fused Feature Reduction and Selection System for Early Lung Cancer Detection

• Published in: 2024 Ninth International Conference on Science Technology Engineering and Mathematics (ICONSTEM) pp. 1 - 8
• Main Authors: Lasrado, Suman Antony, Babu, G N K Suresh
• Format: Conference Proceeding
• Language: English
• Published: IEEE 04.04.2024
• Subjects: Feature and Attribute Reduction; Feature Selection; Information Gain; Linear Discriminant Analysis; Logistic regression; Lung cancer; Lung Cancer Prediction; Machine learning; Machine learning algorithms; Medical services; Predictive models; Runtime
• DOI: 10.1109/ICONSTEM60960.2024.10568801

The research presents ainnovative approach for the speedy detection of lung cancer employing a hybrid Feature Reduction and Feature Selection (FRFS) system. This methodology combines the power of Information Gain (IG) and Linear Discriminant Analysis (LDA) to identify and prioritize informative attributes while filtering out redundant and noisy data elements. Using the UC Irvine Machine Learning Repository dataset, extensive experimentation demonstrates the superior performance of the FRFS system compared to baseline algorithms like Logistic Regression (LR), Decision Tree and Naive Bayes. The results reveal significant enhancements in runtime efficiency, cross-validation precision, and Area under the Curve (AUC) of the ROC. Moreover, the FRFS system enhances interpretability by highlighting key data patterns and feature importance, offering clinicians and researchers actionable insights for informed decision-making in lung cancer prediction. Thrureduction of the dimensions of the attributearea and selecting the most relevant attributes, the FRFS method streamlines the predictive modeling process and improves model generalization. This scalable and efficient solution holds promise in real-world clinical settings, where early detection is paramount for improving patient outcomes and facilitating timely intervention strategies in combating lung cancer. The findings underscore the potential of the proposed FRFS system as a valuable tool in the fight against lung cancer, providing a robust framework for integrating machine learning techniques into clinical practice. By harnessing the collective power of feature reduction and selection, this approach empowers healthcare professionals with actionable insights derived from complex medical datasets. Moving forward, extremeinvestigation and justificationanalyses are affirmed to evaluate the generalizing and scalabiling of the FRFS system across diverse patient populations and healthcare settings.