Text Mining for Customer Sentiment Using Naive Bayes and SMOTE Methods on TokopediaCare Twitter

Abstract

At this time, buying and selling online has become part of the lives of the Indonesian people and the world, especially during the pandemic, marketplace users are increasing and slowly replacing traditional markets. Tokopedia as one of the largest marketplaces in Indonesia has the largest users in the 3^rd quarter of 2019. Customer complaints to Tokopedia services can be submitted through Social Media such as Twitter and also other media. Complaints submitted via Twitter to TokopediaCare are still manually identified by Tokopedia customer service so it takes a long time to respond to customer complaints because customer services need so much time to classify where is a complaint or not complaint tweeted. Text mining is used to process customer complaint data through text or sentences submitted by tweets using the Naïve Bayes method and the Synthetic Minority Oversampling Technique Method (SMOTE) feature for the implementation of machine learning can help identify the classification of complaints submitted via Twitter automatically. The use of the Naive Bayes method is added with the Syntethic Minority Oversampling Method feature which is considered better for generating predictions on tweets submitted by customers..