Title: A cyberstalking-free global network with artificial intelligence approach
Authors: Nureni Ayofe Azeez; Odejinmi Oluwatobi Samuel
Addresses: Department of Computer Sciences, University of Lagos, Nigeria ' Department of Computer Sciences, University of Lagos, Nigeria
Abstract: Cyber harassment is a cybercrime that has posed a great danger to social media users. This work aims at comparing the traditional classifiers and deep learning in detecting cyber harassment. Seven machine learning algorithms - Bernoulli NB, decision tree, Gaussian NB, isolation forest, K nearest neighbour - KNN, random forest and support vector machine were chosen from traditional machine learning algorithms while generalised regression neural network, long short-term memory, multilayer perceptron neural network, radial basis neural network were chosen from deep learning models. With dataset 1, support vector machine and generalised regression neural network had the highest accuracy of 0.9257 and 0.9280, respectively. For dataset 2, random forest and long-term short memory had the highest accuracy of 0.9369 and 0.9978, respectively. For dataset 3, isolation forest and multilayer perceptron neural network had the highest accuracy of 0.6113 and 0.8165, respectively. Finally, the ensemble results yielded an accuracy of 0.9146.
Keywords: cyber-harassment; deep-learning; cybercrime; machine-learning; algorithms; metrics; ensemble.
DOI: 10.1504/IJICS.2023.131096
International Journal of Information and Computer Security, 2023 Vol.21 No.1/2, pp.82 - 108
Received: 22 Jun 2021
Accepted: 22 Oct 2021
Published online: 26 May 2023 *