Data Augmentation to Improve Fake Account Detection Using ANN

The goal of this study is to create a model that can effectively predict whether a social media account is fake. The model is implemented as a web-based interface to give a user-friendly output for managing a large number of profiles on social networking sites. The suggested methodology is believed to be adaptable to a variety of social media sites, with particular emphasis on Instagram. To achieve this purpose, the paper proposes a system to improve the accuracy of detecting fake social media profiles using augmentation approaches such as flipping, jitter, and feature scaling. The performance of the model is validated using several measures, such as F1-score, accuracy, precision, and recall. The study aims to create a model to recognize whether a social media account is fake or not and is implemented as a web-based interface to provide user-friendly output for social networking sites. The methodology followed in this work involves data exploration, data normalization, data augmentation, the development of ANN model, cross-validation, and testing. The model architecture is made up of four fully connected layers. The Adam optimizer is used to compile the model, a popular neural network optimisation algorithm, the categorical cross-entropy loss function, and the accuracy metric. The data set is split into 80:20 ratio for training and testing data set, respectively. The model's accuracy is 93%, and the validation loss is reduced by 10% by doing several augmentation techniques.