An Efficient Optimization of Multimodal Web Page Genre Classification Based on Objects Using LR-YoloV4 and (BM)2-CWRNN Deep Learning Techniques

Abstract

Web data mining has emerged as a convenient and crucial platform for extracting valuable data. In order to upload and download data, users prefer to use the World Wide Web. Therefore, an alternative way is offered by the web classification for supporting effective information retrieval on the Web multimedia data. In this study introduces a video analysis that involves selecting representative frames from a video sequence. Manhattan distance, also known as taxicab distance, is one of the distance metrics used in keyframe extraction. The video quality measure involves comparing the content of the video ad to a reference, such as a non-advertisement video or an ideal ad. SSIM quantifies the structural similarity between the reference and the ad in terms of luminance, contrast, and structure. To identify and categorize objects in video or image, often bounding boxes are drawn around the detected objects. The purpose of YOLOv4 is to design an object detector that operates efficiently in producing systems and can be easily trained and used. The Blue Monkey (BM) algorithm is a novel optimization metaheuristic algorithm that is inspired by the efficient performance of blue monkey swarms in nature to enhance video quality. The various machine learning classifiers were chosen for classification, named BM2-CWRNN. The extracted features from the video, the web pages are considerably categorized by the classifier as per their corresponding domain. The publicly accessible Web classification URL datasets are utilized. The results attained the proposed CWRNN are contrasted with the Brownian motion algorithms. The experimental results indicated that the classification accuracy is higher. The accuracy rates are attained via the proposed BM2-CWRNN and the web pages are effectively classified consistent with their classes.