Optimized generative adversarial network with fractional calculus based feature fusion using Twitter stream for spam detection

Abstract

ABSTRACT The social networks continue to augment their popularity due to the increased usage of the Internet. The people become connected using social media like Facebook and Twitter. This has increased impulsive communication, namely, spam and is utilized in accumulating information of an individual or marketing to cause offense against people. Spam detection in Twitter is a major issue because of small text and elevated language inconsistency in social media. Thus, it is imperative to devise a spam detection model that poses the ability to detect spam messages using Twitter data. This paper devises a novel spam detection model using a stream of Twitter data. Here, the data transformation is done on the input data using Yeo-Jhonson (YJ) transformation for making the data suitable for processing. The feature fusion is performed using Renyi entropy and Deep Belief Network (DBN). Moreover, the spam detection is performed using the Generative Adversial Network (GAN), which is trained by the proposed Conditional Autoregressive Value at Risk-Sail Fish (CAViaR-SF) algorithm. The proposed CAViaR-SF algorithm is devised by integrating Sail Fish optimizer (SFO) and Conditional Autoregressive Value at Risk (CAViaR) algorithm. The proposed CAViaR-SF offered maximal precision of 97.3%, recall of 99.2%, and F-measure of 98.2%.