A Deep Learning Approach for Twitter Bot Detection

Twitter, a social networking platform allows users to convey their ideas. on a wide range of topics, including politics, sports, the stock market, and entertainment. It has a big influence on how people think. A bot on Twitter sends spam messages. As a result, detecting bots aids in spam detection. In this paper, the detection of twitter bots using Deep Learning methods is addressed. At present, the used models aren't updated with latest datasets and have reduced accuracy and some aren't multilingual. A final classifier, Bot-DenseNet, is built on a dense neural network on combining additional metadata with text encodings. Existing methods consider metadata information or along with some semantic features of text in encoding the user account. It will be trained and then verified using extensive data sets collected from Kaggle and twitter API. Subsequently, comparison between the performance of the Bot-DenseNet and Bag-of-Words model is also analyzed.