Details

Autor(en) / Beteiligte

• Manasa, Pinnapureddy
• Malik, Arun
• Batra, Isha

Titel

Detection of Twitter Spam Using GLoVe Vocabulary Features, Bidirectional LSTM and Convolution Neural Network

Ist Teil von

• SN computer science, 2024-02, Vol.5 (2), p.206, Article 206

Ort / Verlag

Singapore: Springer Nature Singapore

Erscheinungsjahr

2024

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• Twitter spam is used to describe any form of unwanted or unsolicited communications, tweets, or activities that users encounter on the social media platform Twitter. This can include things like spam messages from fake accounts, tweets with links to malicious or scam websites, or automated tweets that are sent out in large quantities. Training a model to detect patterns in tweets indicative of spam. The process typically involves collecting a large dataset of labeled tweets, and then using this dataset to train a machine learning model. The proposed model is divided into two parts that operate on the input dataset, which includes tweets and additional meta-data such as the number of followers and actions. The first part focuses on the text of the tweets, using the GLoVe language model to extract vocabulary features which are then used to detect spam using an LSTM deep learning model. The second part utilizes meta-data from the tweets and additional meta-heuristic features, such as the tweet length and number of question marks, and classifies them using a CNN model. The final decision is determined by combining the results from both the LSTM and CNN models.