부산대학교 도서관

A more efficient method of detecting cyberstalking than Convolutional Neural Networks is the objective of this project, which aims to find a means to make Extreme Gradient Boosting a more effective method. Materials and Methods: We tested XGBoost (N = 10) and CNN (N = 10) using datasets with an alpha value of 0.05 and a g-power score of 80%. These datasets were obtained from a variety of internet sources. In Group 1, the predicted incidence rates are 92.53%, whereas in Group 2, the anticipated incidence rates are 87.46%. During the extraction process, the materials are extracted using the BERT characteristic extraction method. A procedure known as BERT characteristic extraction was utilized in order to extract the samples obtained. Results: When the accuracy was enhanced, the BERT model was able to obtain a maximum accuracy of 90% for the CNN classifier. This was its highest possible accuracy. Through the utilization of a value of p=0.000014 (p<0.05), it was determined that there exists a statistically significant disparity between the two methods. Discussion: Investigations such as this one are absolutely necessary in order to enhance the efficiency and accuracy of approaches for detecting cyberstalking. Conclusion: The Convolutional Neural Network surpasses the Extreme Gradient Boosting algorithm in both groups of samples by utilizing BERT feature extraction and better accuracy within the constraints of the data sources. This is the case when comparing the two methods based on their respective capabilities. [ABSTRACT FROM AUTHOR]