Arabic Sentiment Analysis of Mobile Banking Services Reviews
Safaa M. Khabour, Dheya Mustafa, Qutaibah Althebyan · 2023
Online customer reviews have developed into a significant source of information about a business's performance. Due to shifting consumer expectations and growing internet penetration, the Middle East, especially Jordan, is seeing an increase in the usage of online banking. This propensity toward digital transition has been accelerated by the COVID-19 pandemic. Mobile banking service companies aim to effectively use customers' feedback to identify areas for improvement in customer satisfaction. Although Arabic is becoming one of the most widely used languages on the Internet, only a few studies have focused on Arabic sentiment analysis to date. The present study conducts an extensive emotion mining and sentiment analysis on mobile banking reviews in Arabic, exploiting machine learning, natural language processing, and resampling methods to obtain subjective feedback, determine polarity, and identify customers' feelings in the banking domain. This work walks the reader through detailed steps of cleaning and preprocessing a manually annotated dataset of online banking reviews in Arabic, followed by features extraction and modeling, dataset splitting, and balancing training data. Finally, we examined classification algorithms including Decision Tree (DT), Support Vector Machine (SVM), Logistic Regression (LR), Naïve Bayes (NB), K-Nearest Neighbors (KNN), and Random Forest (RF). Ensemble hard and soft voting ML methods were also evaluated. We achieved a maximum accuracy of about 94% using the LR classifier and a maximum precision of 94% using the LR and hard voting methods. Our methodology will enable banks to acquire insights on how to develop their online presence and meet the demands of customers and stakeholders.