Accounting for 22% of global medical deaths, with a death rate of 57%, Lung Cancer stands being the primary cause of cancer-related mortality in the globe, as per the World Health Organization. Early detection and classification through imaging has proven to be crucial for improving survival rates. The study aims at training and validating existing state of the art models on our dataset of 1,190 CT Scan images, which includes well distributed cases spanning across normal, benign and malignant cases. Utilizing a novel approach of transfer of learning features between ResNet50 to the EfficientNetB3 architectures, the model was assessed using standard evaluating metrices including accuracy, precision, recall and F1 Score. The methodology has been able to enhance the metrics for the predictive model’s overall performance with EfficientNetB3 emerging at the top with superior performance of accuracy 96%, precision 91%, recall 90% and an F1 Score of 92%. To further understand the decisions made by the proposed model, an Explainable AI technique SHAP (SHapely Additive exPlanations) heatmaps was applied to the best-performing model for tumor localization, offering a detailed understanding of decision making process followed by the model. The SHAP visualizations enabled us to identify the most influential features contributing to the classification, highlighting the clinical relevance of this technique. The research signifies the impact of both model selection and the application of Explainable AI for enhancing the interpretability and diagnostic reliability in medical imaging and automated diagnoses.