Analisis Sentimen Pada Perusahaan Penyedia Jasa Logistik J&T Menggunakan Algoritma Multinomial Naive Bayes dan Support Vector Machine

Abstract

Online shopping is a way to a faster and easier process of buying things or needs for people these days. Logistic services are essential in the process of buying things online, for they will be the one who ship the package to the buyer. PT. Global Jet Express or J&T is one of many logistics service provider company that are available in Indonesia. J&T has a Twitter account which is used for communicating with their customers. Opinions that were posted by J&T consumers on Twitter could be used as a data to do sentiment analysis which the purpose is to extract information that are told by people in Twitter about J&T. Data crawling was done for 15.000 tweets that were posted during the period of 4^th to 10^th of July 2022, duplicated tweets and those who has the exact same contents were removed resulting the data reduced to 2500 tweets. Tweets will be divided into two class; positive class and negative class Some classification methods are commonly used in text classification, such as Random Forest, Decision Tree, Naïve Bayes Classifier, Support Vector Machine etc. Data in this research will be classified using Multinomial Naïve Bayes and Support Vector Machine to compare their accuracy, the reason for the comparison is these methods have significant difference in their concept complexity. Multinomial Naïve Bayes classify data by finding the greatest conditional probability value, whilst Support Vector Machine classify data by finding the best hyperplane to divide into two class. Multinomial Naïve Bayes has the accuracy of 72,80% and Support Vector Machine has the accuracy of 82,40%. Based on their accuracy, Support Vector Machine has the best performance in classifying public opinions about J&T on Twitter.