Predicting users' behavior using mouse movement information: an information foraging theory perspective

The prediction of users' behavior is essential for keeping useful information on the web. Previous studies have used mouse cursor information in web usability evaluation and designing user-oriented search interfaces. However, we know fairly to a small extent pertaining to user behavior, specifically clicking and navigating behavior, for prolonged search session illustrating sophisticated search norms. In this study, we perform extensive analysis on a mouse movement activities dataset to capture every users' movement pattern using the effects of information foraging theory (IFT). The mouse cursor movement information dataset includes the timing and positioning information of mouse cursors collected from several users in different sessions. The tasks vary in two dimensions: (1) to determine the interactive elements (i.e., information episodes) of user interaction with the site; (2) adopt these findings to predict users' behavior by exploiting the LSTM model. Our model is developed to find the main patterns of the user's movement on the site and simulate the behavior of users' mouse movement on any website. We validate our approach on a mouse movement dataset with a rich collection of time and position information of mouse pointers in which searchers and websites are annotated by web foragers and information patches, respectively. Our evaluation shows that the proposed IFT-based effects provide an LSTM model a more accurate interpretative exposition of all the patterns in the movement of the users' mouse cursors across the screen.