I. Introduction
With millions of daily visitors and a substantial influence on consumer behavior, attitudes, and opinions, Youtube is the most popular video site on the planet. As a result, video optimization[2] has monetary value, and many businesses [3]use it to gain traction and build interest in their products and services. Because of widespread Internet access and the advent of [4]Web 2.0 services, a vast and ever-increasing amount of online data has been introduced into the digital world. Non-traditional channels have allowed content makers to reach consumers in previously inconceivable numbers. [5]Among the many sorts of content available on the internet, online videos are currently the most popular. Video traffic accounted for around 64% of all Internet traffic in 2014, and this figure is anticipated to climb to 80% by 2019 [1]. K-nearest neighbors could be a simple algorithmic software that maintains all existing instances and categorizes new cases based on their similarity. The algorithmic program k-nearest neighbors is a non-parametric technique for classification and regression (k-NN). In each situation, the input is taken from the k-nearest coaching examples within the feature area. The outcome will alter depending on whether k-NN is used for classification or regression: The output of k-NN classification might be a category membership. A majority of its neighbors judge an object [6], and it is assigned to the category that is most common among its k closest neighbors (k could be a positive whole number, usually small). If k = 1, the object is exclusively assigned to that one category of closest neighbors. A k-NN regression is used to determine the item's property price. This is the average of the costs of k of your closest neighbors. K-NN[7] is a type of example-based learning, also referred to as lazy learning, in which work is only approximated regionally and all calculations are postponed until classification. The k-NN algorithmic program is one of the only ones among machine learning algorithms. Applying weight to the contributions of the neighbors[8], so that the closest neighbors contribute more to the common than the farther away neighbors, is an effective method for both classification and regression.