Using Machine Learning Model To Predict Libyan Telecom Company Customer Satisfaction

Abstract

Sentiment analysis is a field that studies the polarity of opinions from texts and determines whether they are positive, negative, or neutral. Since analyzing a large amount of data manually takes a long time, a machine learning-based system had emerged. In this study, a sentiment analysis system to determine the customer opinions of the three major Libyan telecommunication companies namely: (Libyana, Almadar Aljadid, and Libya Telecom and Technology) is proposed, where customer opinions were collected from Twitter. Several pre-processing and cleaning steps had been applied to the collected corpus to improve the performance of the models. Five machine learning models, namely: support vector machine, logistic regression, naive Bayes, K-nearest neighbor, and decision tree have been applied. An initial experiment showed that most of the models are overfitting due to class imbalance. Followed class balancing step is performed for all companies. The results showed that the support vector machine was the best in predicting the customer sentiment of the Libyana telecom company with an accuracy of 80.67%. The naive Bayes was the best on Almadar Aljadid with an accuracy of 81.19%. In Libya Telecom and Technology, the result showed that the performance of the decision tree was the best at 75%. This study showed that the sentiment of Libyan telecom companies was successfully predicted through content posted on the Twitter social media platform, which might assist those companies in improving their services.