Web Page Classification Using Deep Learning with Text and Image-Based Analysis

Abstract

This study investigates an approach that combines text and image data to classify web pages based on their content. In this methodology, Convolutional Neural Networks (CNN) are used to analyze text data, while the YOLOv8 (You Only Look Once v8) algorithm is utilized for the analysis of image content. In the experiments, 20 web pages have been manually classified and these classifications have served as a benchmark for comparing the results of the model. The final classification of a web page was determined by weighing data from YOLO at 30% and data from CNN at 70%. Experimental results of the study have shown a significant improvement in accuracy with a value of 0.602 for the classification algorithm employing both CNN and YOLO, compared to those using either CNN or YOLO exclusively. © The Institution of Engineering & Technology 2023.