Mining research trends with anomaly detection models: the case of social computing research

We proposed in this study to use anomaly detection models to discover research trends. The application was illustrated by applying a rule-based anomaly detector (WSARE), which was typically used for biosurveillance purpose, in the research trend analysis in social computing research. Based on articles collected from SCI-EXPANDED and CPCI-S databases during 2000 to 2013, we found that the number of social computing studies went up significantly in the past decade, with computer science and engineering among the top important subjects. Followed by China, USA was the largest contributor for studies in this field. According to anomaly detected by the WSARE, social computing research gradually shifted from its traditional fields such as computer science and engineering, to the fields of medical and health, and communication, etc. There was an emerging of various new subjects in recent years, including sentimental analysis, crowd-sourcing and e-health. We applied an interdisciplinary network evolution analysis to track changes in interdisciplinary collaboration, and found that most subject categories closely collaborate with subjects of computer science and engineering. Our study revealed that, anomaly detection models had high potentials in mining hidden research trends and may provided useful tools in the study of forecasting in other fields.