Sentimen Analisis Pengguna Twitter pada Event Flash Sale Menggunakan Algoritma K-NN, Random Forest, dan Naive Bayes

Abstract

There is a sales system called Flash Sale in e-commerce. Basically the concept of a Flash Sale is to offer a lower price and a predetermined time and number of products. The sales system is only held at certain moments, by making cheaper product sales but with a limited time and number of products it will make sales increase because buyer interest will be higher. But apart from all the advantages of course there will be pros and cons. Sentiment analysis on Twitter was chosen because Twitter itself is a social media that allows users to be free to comment or write opinions about anything, including opinions about events that exist in e-commerce today. Thus, this research exists to find out the opinions of existing Twitter users regarding the Flash Sale event held by e-commerce. By using the methodology of three classification algorithms, Naive Bayes, K-Nearest Neighbor and Random Forest in classifying the data to determine the accuracy of the sentiment value of Twitter users in the Flash Sale event. This research takes two data samples from the keywords flash sale and flash shopee, the results accuracy of the implementation of the three classification algorithms are 83.53% Naive Bayes, 82.94% K-NN, 80.59% Random Forest for the keyword flash sale” and 81.48% Naive Bayes, 77.78% K-NN, 74.07% Random Forest for the keyword “flash shopee”. With this, the Naive Bayes Algorithm becomes a recommendation for classifying Sentiment Analysis data with greater accuracy and more stability to be used for large and small data.