Predictive visual analytics of event evolution for user-created context

Abstract

Digital big data provide the vast potential of increasing effectiveness and efficiency for decision making. Since the volume of the data is enormous, the data analysis requires large amount of time and effort. It is more problematic when predictive analysis is necessary for futuristic decision making. For predictive analysis, there have been many studies to forecast future trends spatio-temporally. However, most of studies provide just future tendency per event using graphs or maps without contextual compositive analysis. In this paper, we present a predictive visual analytics system to provide predictive event patterns. We infer the future event evolution by combining contextually similar cases occurring in the past. We utilize social media data to detect interesting abnormal events and match the detected abnormal events within the past news media data to retrieve similar event patterns. Then, we extract future event patterns through compositing contextual relationship among topics included in the similar past patterns. To evaluate our VA system, we demonstrate three use cases in this paper and validate our system with possible predictive story lines. In addition, we present an informal user study and feedback to validate the effectiveness of our system and improve the system in the future. Graphical Abstract text