Convolutional Neural Network Implementation Based on SMOTE and Data Augmentation for Lung Cancer Diagnosis

One of the most fatal cancer kinds is lung cancer; the chance of surviving a patient depends upon the stage of diagnosis. The earliest stage of lung cancer is when it is most treatable. However, lung cancer in its early stages frequently exhibits no clinical symptoms and is difficult to detect. Improvement in the survival rate assistant of Deep learning is helpful. Through this, we will try to acquire the suggested aim. A convolutional neural network is a very powerful deep-learning tool used to classify medical images. The lung cancer detection dataset’s issue with imbalanced classes, however, restricts the classifier’s effectiveness for minority classes. We propose two models SCNN and DACNN for imbalance processing technologies for huge datasets in order to increase the identification rate of classes with few images while maintaining efficiency. SCNN approach combines CNN with Synthetic Minority Over-Sampling Technique (SMOTE) and DACNN approach combines CNN with Data Augmentation. The SMOTE and data augmentation is used to increase the clarity of classes and avoid overfitting in building classification models. The proposed algorithm is tested on IQ-OTH/NCCD dataset. With these combinations, the accuracy of these algorithms is achieved at 99.37% and 98.91%.