인공신경망 기반의 SNS 악성 댓글 탐지 기법

Abstract

As the development of Internet technology and the spread and use of smart mobile devices have expanded widely, the lives of modern people have brought many changes and benefits. It has become possible to acquire various information more easily and quickly, and it is also possible to exchange large amount of information easily, quickly. In addition, social networking systems enable people to communicate in real-time with people who are far away. However, this convenience and benefits as well as social problems are occurring continuously. Many people are suffering from problems such as malicious comments, human attacks, privacy invasion, defamation, etc. through the internet space. In this study, one of the ways to reduce these social problems is the frequent This paper proposes a technique to discriminate, detect, and limit malicious comments made in the past. In this experiment, 809 general comments, 1108 malicious comments, total 1917 comments were collected and analyzed, and the malicious comment detection experiment was conducted with the learned neural network. Experimental results show 90.1% of training sample accuracy and 88.1% of sample accuracy.