Machine learning approach to customer sentiment analysis in twitter airline reviews

¹ Doctoral of Information System Department, Diponegoro University, Semarang, Indonesia
² Physics Department, Science and Math Faculty, Diponegoro University, Semarang, Indonesia
³ Informatics Department, Science and Math Faculty, Diponegoro University, Semarang, Indonesia

^* Corresponding author: eka.pujo@hangtuah.ac.id

Abstract

Customers typically provide both online and physical services they use ratings and reviews. However, the volume of reviews might grow very quickly. The power of machine learning to recognize this kind of data is astounding. Numerous algorithms that could be employed for job of sentiment analysis have been developed to categorize tweets about airline sentiment into positive, neutral, or negative categories, this study compares the effectiveness algorithm for machine learning Naive Bayes (NB), Logistic Regression (LR), Decision Tree (DT), Support Vector Machine (SVM), Adaboost, Extreme Gradient Boosting (XGB), Light Gradient Boosting Machine (LGBM), and Random Forest (RF) dividing the Twitter airline sentiment data into positive, neutral, or negative categories using the TF IDF model. The experiment involved two phases of activity: a classification algorithm utilizing SMOTE and sans SMOTE with Stratified K-Fold CV algorithm. With the RF model, the greatest performance accuracy for SMOTE is 97.56%. Without SMOTE, the RF with a value of 92.21% provides the maximum performance accuracy. The findings demonstrate that SMOTE oversampling can improve sentiment analysis accuracy.