Leveraging Bio-Inspired Algorithms and Ensemble Learning for Robust Twitter Spam Detection

The inadequate sampling technique that focusses on the bio-inspired technique and collaborative learning is proposed for the purpose of addressing class disparities while dealing with difficulties related to Twitter spam identification. Due to the uneven nature of the datasets of various communication platforms, non-spam tweets are more prevalent than spam tweets. As a consequence, machine learning algorithms have a low rate of predicting spam. The system that is being given is made up of PSO-based inadequate sampling, which is used to enhance many of the class examples, and CS-RF, which is created to raise the effectiveness of spam identification. Random inadequate sampling, random excessive sampling, and PSO-based inadequate sampling are the three ways of data collecting that are used in this methodology for the purpose of generating equal information. In order to improve the accuracy of predictions, a process known as collaborative learning is used, which involves the integration of classifiers with larger groups of classifiers. This approach displays exceptional accuracy for spam detection, with results indicating greater precision, quantity, and F-measure. Furthermore, comparison of results with existing methods reveals that this method is superior to the current approaches. The real issue of class imbalance and the capability to identify occurrences of the minority class (spam) are both tackled via the utilisation of data-level sampling methods and cost-sensitive algorithms to solve the challenge