Deep Learning Approach for Identifying and Classifying Cancer Types Using BPSO and GAN

Abstract

One of the most dangerous and dreadful diseases in the world is cancer. In 2020, it was found that over 10 million people will die due to cancer. Main causes of cancer are either due to inherited genetics or environmental and lifestyle factors. The risk of cancer increases significantly with age, their type and many cancers occur more commonly in developed countries. Early detection of cancer plays an important role, as it increases the possibilities of betterment. Latest improvement of artificial intelligence techniques in terms of productivity and precision and also optimization algorithms have largely smoothed out the human genomics study. So, we propose an RNA-Sequence analysis method for identifying and classifying cancer types, which uses Binary Particle Swarm Optimization using Support Vector Machine (BPSO-SVM) as feature selection tool and Generative Adversarial Network (GAN) along with different classical augmentation techniques to avoid overfitting by increasing the size of dataset and a deep learning network as the last phase for classification and thus helps in increasing accuracy.