부산대학교 도서관

The main difference between this research and the Convolutional Neural Network approach is that it aims to improve the accuracy of cyberstalking detection by applying Support Vector Machines algorithm. The datasets are gathered from the different sources on internet and have an alpha value of 0.05. The g-power value of both SVM and CNN is 80%. According to the analysis, group 1 and Group2 are projected to have incidence rate of 92.53%-87.46%. For taking the samples, a BERT feature extraction technique is employed. Using the BERT model for feature extraction, a maximum accuracy of ninety percent is being achieved by utilizing the CNN classifier. The results indicated that there is a significant difference between the two algorithms based on p=0.083 and (p< 5). This research is necessary to improve the accuracy of cyberstalking detection systems and ensure their success in practical settings. Despite the limitations on data sources, Convolutional Neural Network has demonstrated better results in both sample groups with BERT feature extraction and a higher level of accuracy compared to that provided by SVM. For this experimental arrangement one needs Ryzen 3 CPU eight gigabytes (8 GB) random access memory(RAM) and fifty gigabyte hard disk space To perform the technical procedure, MATLAB 2014.rar and any version of Windows are needed to be used.