Sentiment Analysis and classification for Software as a Service Reviews

With the rapid growth of cloud services, there has been a significant increase in the number of online consumer reviews and opinions on these services on different social media platforms. These reviews are a source of valuable information in regard to cloud market position and cloud consumer satisfaction. This study explores cloud consumers' reviews that reflect the user's experience with Software as a Service (SaaS) applications. The reviews were collected from different web portals, and around 4000 online reviews were analysed using sentiment analysis to identify the polarity of each review, that is, whether the sentiment being expressed is positive, negative, or neutral. Also, this research develops a model for predicting the sentiment of Software as a Service consumers' reviews using a supervised learning machine called a support vector machine (SVM). The sentiment results show that 62% of the reviews are positive which indicates that consumers are most likely satisfied with SaaS services. The results show that the prediction accuracy of the SVM-based Binary Occurrence approach (3-fold cross-validation testing) is 92.30%, indicating it performs better in determining sentiment compared with other approaches (Term Occurrences, TFIDF). This work also provides valuable insight into online SaaS reviews and offers the research community the first SaaS polarity dataset.