International Journal of English Literature and Social Sciences Vol-9, Issue-6; Nov-Dec, 2024 Peer-Reviewed Journal Journal Home Page Available: https://ijels.com/ Journal DOI: 10.22161/ijels IJELS-2024, 9(6), (ISSN: 2456-7620) (Int. J of Eng. Lit. and Soc. Sci.) https://dx.doi.org/10.22161/ijels.96.64 386 YouTube’s Two-Stage Deep Learning Framework for Video Recommendations Sameena Begum 1 , Swati Chauhan 2 , Sana Sarwar 3 , Dr Gulnaz Fatma 4 1 Language Instructor, Dept of Foreign Languages, College of Arts and Humanities, Jazan University, KSA sbmohammed@jazanu.edu.sa 2 Language Instructor, Dept of Foreign Languages, College of Arts and Humanities, Jazan University, KSA schauhan@jazanu.edu.sa 3 Language Instructor, College of Science for Girls, Jazan University, Jazan KSA ssarwar@jazanu.edu.sa 4 Language Instructor, University College, Al Ardah, Jazan University, Jazan, Saudi Arabia gulnaz.fatima15@gmail.com Received: 23 Nov 2024; Received in revised form: 20 Dec 2024; Accepted: 24 Dec 2024; Available online: 31 Dec 2024 ©2024 The Author(s). Published by Infogain Publication. This is an open-access article under the CC BY license (https://creativecommons.org/licenses/by/4.0/). Abstract— One of the most popular and technologically advanced commercial recommendation systems is YouTube, which makes it one of the most popular. In addition to providing an overview of the system, it shows the extraordinary advances in efficiency that have been made possible by advanced learning techniques. The standard "two-stage" architecture is utilized in this research for the purpose of isolating and obtaining related material. First, a comprehensive model for the generation of candidates should be provided, and then a model for ranking applicants should be outlined. Both of these should be done in the order that they are stated. Important consideration should be given to the sequence in which these two activities should be finished. Additionally, it provides instruction that can be put into practice as well as useful insights that have been gathered from the process of designing, iterating, and maintaining a large- scale recommendation system that has a substantial impact on the experience of the user. In addition, it offers training that can be put into practice. Keywords— Deep Learning, Framework, Ranking, Creating, Iterating. I. INTRODUCTION YouTube has gained widespread recognition as the most prominent global platform for the creation, dissemination, and exploration of material that is based on video technology. By pulling from a vast library of videos, the recommendation system on YouTube plays a vital part in allowing the discovery of material that is specifically customized to the preferences of over one billion users. The purpose of this research is to investigate the major impact that machine learning has had on the video rating system that YouTube has utilized over the past few years. A representation of the suggestions that appear on the home screen of the YouTube application for mobile devices is shown in Figure 1. There are three primary areas in which it is challenging to create recommendations for videos on YouTube: A great number of the existing recommendation algorithms, which have been demonstrated to perform effectively in limited circumstances, require increased scalability in order to be helpful for their requirements. Scattered learning algorithms and speedy, dependable portion systems are essential for YouTube because of the enormous user base and material collecting that makes it possible for YouTube to function. Considering that YouTube uploads a growing number of videos each and every second, the material of the website is always evolving. When it comes to accurately processing freshly delivered content and user actions, the efficiency of the reference system needs to be sufficient. Maintaining a sufficient amount of fresh information at all times.