Automated Spam Detection Using ECSA-Based Feature Selection with BGRN Classifier in Soft Computing Applications

Soft computing techniques are widely used due to their robustness and tolerance for imprecision and uncertainty. They are employed in various fields like control systems, prediction systems, enhancements in robotics, risk management, and more. The flexibility and adaptability of soft computing make it an indispensable area of study, especially as systems and datasets continue to grow in complexity and size. With the use of cutting-edge technology, cybercriminals are increasingly able to send very convincing spam that can compromise sensitive data. The employment of cutting-edge methods to prevent spam has become crucial to the success of this work. Feature subset selection is a crucial step in many machine learning, data mining, and classification workflows. Feature subset selection aims to reduce the complexity of the issue while preserving the most discriminating data necessary for precise classification. The creation of a multi-objective feature extractor and a bidirectional gated recurrent network (BGRN) classifier for Twitter spam detection is seen as the key contribution here. Extracting features from text datasets using mini-batch K-Means normalized mutual information for improving the classification accuracy. The enhanced crowd search algorithm (ECSA) is used to pick features due to the seemingly large length of the features and to decrease the training complexity. In addition, the suggested ECSA is employed to enhance a unique BGRN for application in BGRN spam detection. The BGRN model we suggested was 98% accurate. Comparing the proposed model to already-existing methods, we find that it significantly increases spam detection accuracy.