Lung cancer (LC) is one of the fatal diseases for which early diagnosis of the diseased cell is of utmost importance in order to lower the possibility of death. Computed Tomography (CT) scans are a major imaging modality for diagnosis of LC. Nowadays, for accurate and fast detection of cancerous cells from huge databases of CT images, computational methods especially machine learning (ML) and deep learning (DL) algorithms are leveraged with excellent success. In this work, we attempted to get one lung cancer classification (LCC) predictor that uses transfer learning (TL) based on ResNet50. The experiments were performed on ‘Chest CT-scan Images Dataset’ that has CT scans of normal cell and three variants of LC. In this work, we have attempted to classify the CT scans of LC images using different convolutional neural network (CNN), e.g. ResNet50, DenseNet201 and EfficientNetB4 based TL. The experimental results demonstrate that by fine-tuning the hyperparameters and by changing the weights of the dropout layer and patience in early stopping, ResNet50 with Adam optimizer, provides superior performance compared to DenseNet201 and EfficientNetB4. The proposed predictor was able to obtain encouraging accuracy of 91.87% which provides a potential strategy for diagnosis of LC variants.