HARMONY SEARCH-BASED XCEPTION MODEL FOR ENHANCED LUNG CANCER PREDICTION

Lung cancer continues to be a major cause of cancer-related deaths globally, highlighting the importance of precise and effective diagnostic methods. Deep learning models, such as Xception, have shown promise in medical image classification, but their performance depends heavily on optimal hyperparameter selection. Traditional tuning methods often fail to achieve the best results efficiently, necessitating advanced optimization techniques. This study integrates the Harmony Search (HS) algorithm with the Xception model to optimize hyperparameters for improved lung cancer classification. The proposed HSXception model is evaluated against AlexNet, Inception V3, SqueezeNet, and standard Xception on a dataset containing four lung cancer categories. The models are assessed using multiple performance metrics. The experimental results demonstrate that HS-Xception outperforms all competing models, achieving a 97.18% accuracy, 97.37% precision, 97.33% recall, and a 99.95% AUC-ROC score, significantly improving upon standard Xception (93.22% accuracy). The optimization enhances model efficiency, reducing log loss and improving classification reliability across all lung cancer types. The integration of the Harmony Search algorithm with Xception enhances hyperparameter tuning, significantly improving lung cancer prediction accuracy. This study showcases the potential of metaheuristic optimization in Deep Learning (DL)-based medical image analysis, contributing to the development of more reliable and efficient diagnostic systems.