Scholarlink

Due to the complexity of online communication, this study presents a novel and effective approach for cyberbullying detection. Natural Language Processing (NLP) methodologies are applied and performance results are evaluated of logistic regression and decision tree model on the dataset of tweets. Research has introduced several methods such as sentiment analysis, N-gram analysis, term frequency-inverse document frequency, and profanity detection to improve the accuracy in planning for cyberbullying detection. Before or after tuning, the comparison between logistic regression and decision tree model indicates that the value of logistic regression is much better than decision tree model in almost all of the metrics. It can also be observed that the tuned logistic regression model gives an F1-score of 0.87, accuracy of 0.87, precision of 0.93, and recall of 0.82. The tuned decision tree model achieved a 0.70 F1-score, 0.70 accuracy, 0.79 precision, and 0.65 recalls. Results delineate the logistic regression's effectiveness, notably driven by optimized hyperparameters (C = 1, solver = ‘liblinear’, penalty = ‘l2’), facilitating improved cyberbullying detection in digital spaces. Combining machine learning, NLP, and sentiment analysis forms a well-rounded approach for recognizing and eliminating cyberbullying.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing Cyberbullying Detection: A Multimodal Approach Leveraging Sentiment Analysis and N-gram Modeling

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Pratap Singh Barth,
  • Kunal Bhushan Ranga,
  • Harvir Singh

摘要

Due to the complexity of online communication, this study presents a novel and effective approach for cyberbullying detection. Natural Language Processing (NLP) methodologies are applied and performance results are evaluated of logistic regression and decision tree model on the dataset of tweets. Research has introduced several methods such as sentiment analysis, N-gram analysis, term frequency-inverse document frequency, and profanity detection to improve the accuracy in planning for cyberbullying detection. Before or after tuning, the comparison between logistic regression and decision tree model indicates that the value of logistic regression is much better than decision tree model in almost all of the metrics. It can also be observed that the tuned logistic regression model gives an F1-score of 0.87, accuracy of 0.87, precision of 0.93, and recall of 0.82. The tuned decision tree model achieved a 0.70 F1-score, 0.70 accuracy, 0.79 precision, and 0.65 recalls. Results delineate the logistic regression's effectiveness, notably driven by optimized hyperparameters (C = 1, solver = ‘liblinear’, penalty = ‘l2’), facilitating improved cyberbullying detection in digital spaces. Combining machine learning, NLP, and sentiment analysis forms a well-rounded approach for recognizing and eliminating cyberbullying.