This study discusses the result of sentiment analysis of comments on Youtube videos using the K-Nearest Neighbor method. It contains various opinionsabout video content that have been watched. Opinnios given in comments can be used as an assessment and analyze how the sentiment rises. WhileYouTube only facilitates like and dislike buttons which can be seen from the number of clicks, the comments in a video will be used to analyze a sentiment that appears. Through this research, the system will classify each comment contained in the video, and make categories into positive and negative sentences. Before the classification results are obtained, it will go through several stages such as preprocessing, term weighting, similarity calculations and arrive at the calculation of accurate results. In addition to the calculation accuracy, there are also calculations of precision and recall. In this study using 4 scenarios with differences in the percentage of the amount of testing, training data and the value of K, with the aim of finding the best accuracy. The conclusion based on the results of the study is the large amount of training data, testing data and the value of K affects the accuracy results.The amount of testing data and training data also affects the accuracy search time. The highest accuracy is 92.71% with 3% testing data and 97% training data with k-7. The utilization of this KNN method has a good and accurate performance in the process of classifying YouTube video comments