|Keywords:||Text Visual analytics; Data visualisation; Online reviews classification; Multi-dimensional data visualisation; Visual feature selection|
|Full text PDF:||http://hdl.handle.net/1805/12483|
Indiana University-Purdue University Indianapolis (IUPUI) The purpose of this work is to prove that the visualization is at least as powerfulas the best automatic feature selection algorithms. This is achieved by applyingour visualization technique to the online review classification into fake and genuinereviews. Our technique uses radial chart and color overlaps to explore the bestfeature selection through visualization for classification. Every review is treated as aradial translucent red or blue membrane with its dimensions determining the shapeof the membrane. This work also shows how the dimension ordering and combinationis relevant in the feature selection process. In brief, the whole idea is about givinga structure to each text review based on certain attributes, comparing how differentor how similar the structure of the different or same categories are and highlightingthe key features that contribute to the classification the most. Colors and saturationsaid in the feature selection process. Our visualization technique helps the user getinsights into the high dimensional data by providing means to eliminate the worstfeatures right away, pick some best features without statistical aids, understand thebehavior of the dimensions in different combinations.Advisors/Committee Members: Fang, Shiaofen.