REDUCING DATA VOLUME IN NEWS TOPIC CLASSIFICATION USING DEEP LEARNING

Abstract

The exponential growth of digital news content has posed significant challenges in efficiently processing and classifying large volumes of textual data. News topic classification is essential for organizing, filtering, and retrieving relevant information from extensive news datasets. However, high-dimensional text data increases computational complexity, memory usage, and training time in deep learning models. This project focuses on reducing data volume in news topic classification while preserving classification accuracy through advanced deep learning techniques. The proposed approach incorporates effective text preprocessing and dimensionality reduction methods, including tokenization, stop-word removal, stemming, and feature selection, to eliminate redundant and irrelevant information from news articles. By optimizing the input data size, the system enhances computational efficiency and reduces resource consumption. A deep learning model is then utilized to classify news articles into predefined categories such as politics, sports, technology, and business. Experimental results indicate that data volume reduction significantly improves training speed and overall efficiency with minimal impact on classification performance. The study emphasizes the importance of efficient data preprocessing and feature optimization in large-scale text classification tasks. The proposed system can be effectively applied in news recommendation systems, automated journalism analysis, and large-scale information retrieval applications.