Predictive Visualization

A Data Modeling Approach to Predict Users' Interests With Visualization Systems

Fig: Visual Interface used in our user study

• Visualizations help tackle a class of problems that neither the computer nor the human can solve on their own. The computer performs complex computations and displays a visual representation of the data so that the human can reason and make judgments. However, the computer's role is typically limited once the data is displayed.


• Our research aims to further support the human by predicting their interactions with a visual interface. In our research, we will be developing an algorithm that uses clustering to create models of the data. We will analyze users’ interations to identify the data cluster that best represents users’ interactions or interest.


• To evaluate the algorithm, we will be conducting an online user study and recording users’ mouse clicks as they interact with a crime map. Users’ mouse clicks will be analyzed using our algorithms to predict what features of crime (type, regions etc.) the users are interested in. This research lays the foundation for developing interfaces that would adapt to assist the users in their tasks and thus create predictive visualizations.