Assessing Institutional Performance using Machine Learning on Arabic Facebook Comments

Abstract

Social networks have become increasingly influential in shaping political and governmental decisions in Middle Eastern countries and worldwide. Facebook is considered one of the most popular social media platforms in Iraq. Exploiting such a platform to assess the performance of institutions remains underutilized. This study proposes a model to help institutions, such as the Iraqi Ministry of Justice, evaluate their performance based on sentiment analysis on Facebook. Different machine learning algorithms were used, such as Support Vector Machine (SVM), Logistic Regression (LR), Extreme Gradient Boosting (XGBoost), Naive Bayes (NB), and Random Forest (RF). Extensive experimental analysis was performed using a large dataset extracted from Facebook pages belonging to the Iraqi Ministry of Justice. The results showed that SVM achieved the highest accuracy of 97.774% after retaining certain stop words, which proved to have a significant impact on the accuracy of the algorithms, ensuring the correct classification of comments while preserving the sentence's meaning.