The surge in online media has inundated the public with information, prompting the use of sensational and provocative language to capture attention, worsening the prevalence of online malicious behavior. This study delves into machine learning (ML) and deep learning (DL) techniques to identify and recognize harmful news and toxic comments, aiming to counteract the detrimental impact on public perception. Effective methods for detecting and categorizing malicious content are proposed and discussed, highlighting the differences between ML and DL approaches in combating malicious behavior. The study employs feature selection methods to scrutinize the distinctive feature set and keywords linked to harmful news and toxic comments. The proposed approach yields promising outcomes, achieving a 94% accuracy rate in recognizing toxic comments, a 68% recognition accuracy for harmful news, and an 81% accuracy in classifying malicious behavior content (combining harmful news and toxic comments). By harnessing the capabilities of ML and DL, this research enriches our comprehension of and ability to mitigate malicious behavior in online media. It provides valuable insights into the practical identification and categorization of harmful news and toxic comments, highlighting the unique facets of these advanced computational strategies as they address the pressing challenges of our digital society.

网络媒体的激增使公众信息泛滥，促使人们使用耸人听闻和挑衅性的语言来吸引注意力，从而加剧了网络恶意行为的流行。这项研究深入研究机器学习（ML）和深度学习（DL）技术来识别和识别有害新闻和有毒评论，旨在抵消对公众认知的有害影响。提出并讨论了检测和分类恶意内容的有效方法，强调了机器学习和深度学习方法在打击恶意行为方面的差异。该研究采用特征选择方法来仔细审查与有害新闻和有毒评论相关的独特特征集和关键词。该方法取得了可喜的成果，恶意评论识别准确率达到 94%，有害新闻识别准确率达到 68%，恶意行为内容分类（有害新闻和恶意评论结合）准确率达到 81%。通过利用机器学习和深度学习的功能，这项研究丰富了我们对在线媒体中恶意行为的理解和减轻在线媒体恶意行为的能力。它为有害新闻和有毒评论的实际识别和分类提供了宝贵的见解，突出了这些先进计算策略在解决数字社会紧迫挑战时的独特方面。