Tweet Classification Using Sentiment Analysis Features and TF-IDF Weighting for Improved Flu Trend Detection

Social Networking Sites (SNS) such as Twitter are widely used by users of diverse ages. The rate of the data in SNS has made it become an efficient resource for real-time analysis. Thus, SNS data can effectively be used to track disease outbreaks and provide necessary warnings earlier than official agencies such as the American Center of Disease Control and Prevention. In this study, we show that sentiment analysis features and weighting techniques such as Term Frequency-Inverse Document Frequency (TF-IDF) can improve the accuracy of flu tweet classification. Various machine learning algorithms were evaluated to classify tweets to either flu-related or unrelated and then adopt the one with better accuracy. The results show that the proposed approach is useful for flu disease surveillance models/systems.