Abstract
Deep learning (DL) is the leader in data science, and this has piqued the interest of researchers and businesspeople alike in machine learning. Multiple layers of representational data theories are used in DL's model-building process. Model transfer (MT), convolutional neural networks (CNN), and generative adversarial networks (GAN) are just a few of the main DL approaches that have fundamentally reworked our view of data processing. In fact, DL's processing capacity is astounding when applied to the analysis of pictures, texts, and voices. Evaluation of this data using traditional methods and techniques is hard and unmanageable due to the fast expansion and broad availability of digitalized social media (SM). The solutions provided by DL techniques are predicted to be effective in dealing with these issues. Thus, we consider the pre-built DL approaches that have been implemented with respect to social media analytics (SMA). Instead of focusing on the nuts and bolts of DL, we focus on problem domains that provide significant obstacles to SM and offer suggestions on how to overcome them.