<p>Lung cancer is a serious condition characterized by the abnormal and uncontrollable, growth of lung cells, which can lead to severe health complications and even death. Lung cancer is a critical health condition marked by abnormal cell growth in the lungs often resulting in severe complication or death. Various AI approaches are used for the early detection of lung cancer; however, challenges such as limited dataset availability hinder accurate prediction and segmentation consistency affecting prediction performance. Also, a single classifier is often used for simplicity, ease of implementation, but it suffers from higher susceptibility to over fitting and reduced accuracy. To improve predictive accuracy and robustness, stacking based models combine with multiple classifiers to provide a better solution for early lung cancer diagnosis. Therefore, a stacking-based deep ensemble classifier with novel segmentation model is developed for lung cancer prediction. Initially, input images of lung cancer are collected and pre-processed using Bayesian deep matrix factorization to remove noise. Then, the LightenNet enhances the contrast of the noise-free images. After pre-processing, the Bridging ConvNeXt and U-Net (BCU-Net) segments the tumour region. The segmented image is then input into the stacking based ensemble learning classifier. FragNet and SABNet are the two basic classifier used for extract the features, which are then passed to a Meta classifier called Deep Belief Network (DBN) to predict the lung cancer such as Benign, malignant and normal. The proposed model demonstrated the strong predictive performances achieving 96% accuracy, 91.60% hit rate, 97.30% selectivity and 96.70% NPV. By using this proposed approach enables the early prediction of even small residual tumours, suggesting that patients could benefit from a more precise and aggressive intervention strategy.</p>