Performance of the K-Nearest Neighbors Method on Analysis of Social Media Sentiment

Abstract

Each user who interacts with the Internet and information technology can provide feedback on the specific application. One of the applications which were used as social media is youtube. Various comments are given so that it becomes a challenge for the organizers. In this study, the concept of data mining was needed through the K-Nearest Neighbor method as a tool for classifying in addition to investigating comments that have the potential to be sentimental. There are three factors that are observed as input data, namely Services Quality, Information Quality, and Responsibility when the dataset is collected from social media applications. The initial phase of analysis  is to extract the datasets from youtube then carried out pre-processing to the analysis phase.  As the result shows that the method which was proposed  is able to describe the accuracy rate of up to 88% through the confusion matrix technique. Therefore, the performance of K-Nearest Neighbor has provided a classification with positive and negative sentiment analysis classes.