Visual data mining in a Q&A based social media website

Abstract

Data mining methods and technologies have been applied to different social media environments but seldom applied to narrative information based Q&A sites. This paper aimed to employ visual data mining techniques to address health care consumer terms use behavior in the Yahoo!Answers. Three months of data on the topic of diabetes in the health category of Yahoo!Answers were collected and analyzed. Terms from the collected data set were processed, validated, and classified. Both Multi-dimensional Scaling and Social Network Analysis visualization methods were employed to visualize the relationships of terms from related categories ('Complication & Related Disease' and 'Medication'; 'Complication & Related Disease' and 'Sign & Symptom'). Patterns and knowledge were revealed and discovered from the mapping of terms such as acarbose might cause a side effect of hives, antidepressant may increase the risk of developing diabetes, there is a connection between imbalance and birthdefects, etc. The results of this study can be of benefit to both health consumers and medical professionals.