Sentiment Analysis of Twitter Data for Saudi Universities

Abstract

With the tremendous increase in web technologies, many people are conveying and expressing their feelings through social media. These data collections, called big data, can be of high value to institutions, governments, and researchers if analyzed and interpreted as well. Most of the studies on sentiment analysis have been performed for the English language and few of these works focus on Arabic. Further, these studies rarely focus on colloquialisms, and no research was carried out on Twitter data related to Saudi universities. Therefore, this paper aims to develop a sentiment analysis system for analyzing Tweets generated by Saudi Twitter users about Saudi universities. There are two proposed different models used for Twitter-based sentiment analysis. The first model begins with Tweet collection followed by preprocessing. Latent Dirichlet Allocation, used here, eliminated Tweets that were not expressing sentiments; therefore, the document will be reduced, which affects system accuracy. Finally, Support-Vector Machine (SVM) chosen to perform Tweet classification. The second model is based on using different classifier models, such as Naïve Bayes (NB), Stochastic Gradient Descent (SGD), Sequential Minimal Optimization (SMO), K-nearest neighbors Classifier (KNN), Random Forest, Multilayer Perceptrons using Deeplearning, (MPD), and Sentiment Score Calculation (SSC).