Accuracy enhancement in Detection of Malicious Social Bots Using Reinforcement Learning Technique with URL Features in Twitter Network Through Convolutional Neural Network over K -Nearest Neighbors
Mukul Kumar, P. Shyamala Bharathi, G. Sajiv · 2024
This research aims to use Twitter's network's universal resource locator properties, in conjunction with a convolutional neural network (CNN) instead of a K-nearest neighbor (KNN) algorithm, to more effectively identify harmful social bots via reinforcement learning. A reinforcement learning method is used in conjunction with the URL characteristics of the Twitter service, utilizing a Novel CNN over KNN, to identify the performance enhancement of unauthorized social bots. This study used a pre-test power analysis of 80%, with a total of 20 participants divided evenly between the two groups (10 in each). We employed a CNN over KNN reinforcement learning method to detect malicious social bots on Twitter. The network achieved an efficiency of 92.70%, and when we compared it to KNN, we found that it was 87.18% accurate. A p-value of less than 0.05 indicates statistical significance in the accuracy values among the two algorithms, as shown by the independent sample (IS) test. Compared to the KNN method, the CNN algorithm clearly outperforms it, according to the findings.