Prediction of disease is essential for identifying the risk factor. Lung cancer is a dangerous disease which has to be detected in initial stage. Based on images the risk of the disease is identified but analysing for every patient is tuff. By utilizing DL techniques, the prediction can be analyzed in short period of time. For predicting images CNN is the most efficient methodology for its layer structure. The proposed model to increase the efficiency of the system it has generated augmented images using GAN approach because the performance and generalizability of the trained models may be impacted by biases introduced by GANs into the generated data. Thorough validation and assessment are required to reduce these hazards.Class imbalances are typical in medical datasets, where there may be considerable differences in the quantity of samples for distinct classes. By creating artificial examples for underrepresented classes, GANs can assist in balancing the dataset.The synthetic images are passed as input to the fine-tuned AlexaNet then it has extracted the features using vector integrated feature elimination. Since every layer in AlexNet's feature extraction process captures a distinct level of abstraction in the image, the features are generally interpretable. Understanding which characteristics are crucial for identifying malignant from non-cancerous areas in lung imaging can be aided by this. Finally, the model detects the cancer in lungs using the non-linear SVM. Hence the prediction was performed through validation techniques, ROC, AUC was also performed.