Lung cancer is one of the most prevalent illnesses in the world and a leading contributor to rising death rates. Early detection of cancer enables treatment. Early-stage lung cancer classification is difficult because of varied clinical presentations, leading to increased processing time and resource demands for clinicians. Deep learning techniques have been used extensively in a number of medical professions in recent years to improve early diagnosis and treatment stages. In cases of various malignant tumors, such as lung cancer, when time is of the essence for promptly recognizing the patient's condition. A novel deep learning-based LUNCERDEN model has been presented for lung cancer classification in order to address these issues. Lung cancer can be detected from CT scans using this process. Initially the input image CT is gathered from the available datasets. Adaptive median filters are used as a preprocessing technique to lower noise and enhance input image quality. For feature extraction, the preprocessed image is fed into Inception ResNet. Finally, a dense neural network is employed to categorize the many forms of lung cancer, including lung nodules, small-cell lung cancer, and normal lung cancer. The proposed classification accuracy is 98.17%, which is extremely accurate. The proposed LUNCRDEN model improves overall accuracy by 4.17%, 7.32%, and 0.97% in comparison to support vector machines, GoogleNet, and convolutional neural networks, respectively.