Optimized Transfer Learning and Machine Learning Hybrid Approach for Early Lung Cancer Diagnosis

Lung cancer remains one of the most prevalent and life-threatening diseases globally, emphasizing the need for early and precise detection to improve patient outcomes. Traditional deep learning models, despite their success in medical image analysis, often encounter overfitting, high computational demands, and limited interpretability, while classical ML models struggle to automatically extract high-level spatial features from medical images. To bridge this gap, this study presents a hybrid Deep DL+ML framework for lung cancer detection using computed tomography (CT) scan images. The proposed methodology integrates three well-established transfer learning models VGG16, InceptionV3, and ResNet50— to extract rich deep feature representations, which are subsequently classified using six machine learning classifiers: SVM, Random Forest, Gradient Boosting, XGBoost, LightGBM, and k-NN. The extracted feature embeddings undergo optimization and fine-tuning to enhance classification accuracy while addressing challenges such as data imbalance and computational inefficiencies. Experimental evaluations on the Kaggle lung cancer dataset demonstrate that the hybrid DL+ML approach significantly outperforms traditional standalone CNN classifiers. Among the evaluated models, ResNet50 + XGBoost achieved the highest accuracy of 96.8%, highlighting the effectiveness of combining deep feature extraction with advanced ML classification techniques. This study underscores the clinical potential of hybrid AI-driven lung cancer diagnosis, offering an interpretable, computationally efficient, and high-performance framework for early detection in medical imaging.