Predicting User's Intentions With Visualization Systems
Tackling problems through visualization requires humans to draw inferences from complex visual representations. However, the computer's role is typically limited once the data are displayed. Our research aims to further support humans by predicting their interactions. In our research, the users interact with a crime map. Users’ mouse clicks are analyzed using Hidden Markov Model and clustering algorithms to predict what features of crime (type, region etc.) they are interested in. Based on the predictions, the interface would adapt to assist the users in their tasks and thus create predictive visualizations.