Scholarlink

In order to enhance the classification models, this research presents a unique method of pre-processing the data that includes text clustering. The dataset used in this study consists of news articles that have been labelled as “real” or “fake”. The two primary elements of the method under discussion are text data pre-processing and data modelling. Initially, a sparse matrix is created by applying term frequency-inverse document frequency (TF-IDF) vectorization to the text data from the news articles. A piece of news in vectorized form is represented by each row in the matrix. Agglomerative clustering is applied based on these vectors, and as an extra feature, the data is divided into two new clusters. The model is trained using both the modified and the basic datasets alternately in the second stage, and the accuracy scores are compared. With accuracy, precision, and an F1 score of 98%, the multinomial classifier utilizing TF-IDF after stemming produces the best results according to the experiments.

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

A Clustering Ensemble Approach to Improving Text Classification Adding Anti-dictionary

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Titu Singh Arora,
  • Amit Sharma,
  • Punita Panwar,
  • Vijay Singh Rathore

摘要

In order to enhance the classification models, this research presents a unique method of pre-processing the data that includes text clustering. The dataset used in this study consists of news articles that have been labelled as “real” or “fake”. The two primary elements of the method under discussion are text data pre-processing and data modelling. Initially, a sparse matrix is created by applying term frequency-inverse document frequency (TF-IDF) vectorization to the text data from the news articles. A piece of news in vectorized form is represented by each row in the matrix. Agglomerative clustering is applied based on these vectors, and as an extra feature, the data is divided into two new clusters. The model is trained using both the modified and the basic datasets alternately in the second stage, and the accuracy scores are compared. With accuracy, precision, and an F1 score of 98%, the multinomial classifier utilizing TF-IDF after stemming produces the best results according to the experiments.