Anomaly detection through enhanced sentiment analysis on social media data

• Main Authors: WANG, Zhaoxia, JOO, Victor, TONG, Chuan, XIN, Xin, CHIN, Hoong Chor
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2014
• Subjects: Sentiment classification; Social media; Twitter; Anomaly detection; Enhanced sentiment analysis; Machine-learning; Pattern classification; Numerical Analysis and Scientific Computing; Social Media
• Online Access: https://ink.library.smu.edu.sg/sis_research/5567; https://ink.library.smu.edu.sg/context/sis_research/article/6570/viewcontent/AnomalyDetection_2014_av.pdf
• Institution: Singapore Management University

Anomaly detection in sentiment analysis refers to detecting abnormal opinions, sentiment patterns or special temporal aspects of such patterns in a collection of data. The anomalies detected may be due to sudden sentiment changes hidden in large amounts of text. If these anomalies are undetected or poorly managed, the consequences may be severe, e.g. A business whose customers reveal negative sentiments and will no longer support the establishment. Social media platforms, such as Twitter, provide a vast source of information, which includes user feedback, opinion and information on most issues. Many organizations also leverage social media platforms to publish information about events, products, services, policies and other topics frequently. Thus, analyzing social media data to identify abnormal events in a timely manner is a beneficial topic. It will enable the businesses and government organizations to intervene early or adopt proper strategies if needed. However, it is also a challenge due to the diversity and size of social media data. In this study, we survey existing anomaly analysis as well as sentiment analysis methods and analyze their limitations and challenges. To tackle the challenges, an enhanced sentiment classification method is proposed and discussed. We study the possibility of employing the proposed method to perform anomaly detection through sentiment analysis on social media data. We tested the applicability and robustness of the method through sentiment analysis on tweet data. The results demonstrate the capabilities of the proposed method and provide meaningful insights into this research area.