Sentiment Analysis of Expedition Customer Satisfaction using BiGRU and BiLSTM

Abstract

The occurrence of a pandemic caused behavioral changes that occurred in Indonesian society, especially in increasing interest in online purchases. The increased purchases of goods increased the volume of four expeditions, namely: JNE, JNT Express, Sicepat, and Anteraja. To find out the customer satisfaction of the users of the four expeditions automatically, sentiment analysis was conducted based on the thousand tweet data from the opinions of expedition users in three-class categories, which are positive, negative, and neutral. Two deep learning methods were used to analyze the sentiment of expedition customer satisfaction: BiGRU and BiLSTM. The activities conducted during the sentiment analysis were crawling, preprocessing, data labeling, modeling, and evaluation. The performance evaluation results of the two methods use an accuracy matrix over 1,217 test data. The BiGRU method produces an accuracy performance of 71.5% and the BiLSTM method produces an accuracy performance of 66.5%.