Automated Spam Detection Using Stochastic Gradient Descent with Self-Attentive Deep Learning Model
Pinnapureddy Manasa, Arun Malik, Isha Batra · 2022
Since the usage of the Internet is rising, individuals were connected virtually through social networking sites like Facebook, Instagram, Twitter, and so on. This has resulted in surge in the spread of unwanted messages called spam that is utilized for accumulating personal data, marketing, or to offend individuals. As a result, it becomes essential to have a stronger spam detection method that might thwart this type of message. Spam detection is the most important machine learning-oriented application over the past few years. Simultaneously, spam detection on noisy platforms like Twitter which remains a challenge because of high variability and short text in the language used on social networking platforms. To resolve these issues, this paper presents an automated spam detection using stochastic gradient descent with deep learning (ASD-SGDDL) technique. The presented method had a focus towards the detection of spam in the Twitter data. The proposed ASD-SGDDL model pre-processes the tweets in different ways such as filtering, tokenization, stop word removal, and n-gram construction. Followed by, the ASD-SGDDL model utilizes self-attentive deep learning (SADL) method for spam detection and classification. For improving the performance of detection in the presented method, the hyperparameter tuning process can be carried out by utilizing the SGD model. The experimental validation of the proposed method is examined on a benchmark dataset and the outcomes are examined interms of different measures. A wide ranging comparison study stated the enhanced performance of the ASD-SGDDL method to other existing models.