Reinforcement Learning with URL Features in Twitter Network to Detect Malicious Social Bots using Random Forest in Comparison with Convolutional Neural Network to Improve Accuracy
Ram Kumar. M, P. Shyamala Bharathi, S. Rimlon Shibi · 2023
AIM: The objective of the project is to improve the efficiency of identifying dangerous social bots in the Twitter network by developing uniform resource locator characteristics and reinforcement learning. Using random forest in comparison with CNN.Malicious social bot detection accuracy is being improved using the reinforcement learning technique. In contrast to CNN, the Twitter Network uses random forest for URL features. Pre-test power analysis was conducted in this instance using 80%, and a sample size of 20 people was used for each of the two groups, with 10 people in each group. Using a reinforcement learning technique, malicious social bot detection Convolution neural network accuracy was compared to URL features in the Twitter network, which used a random forest with a 99.86% accuracy rate and a convolution neural network with an accuracy rate of 95.89%. The independent sample test yields a statistically 2-tailed significant difference between the two classifiers' accuracy of 0.000(p<0.05).To draw the conclusion that the CNN system is greatly outperformed by the Novel Random Forest algorithm.