Modeling online user product interest for recommender systems and ergonomics studies

Summary This paper presents use of a tool built to monitor human‐website interaction without the need for eye tracking. Behavior indicators calculated from data collected by the tool can be used for many purposes such as web ergonomics enhancement, content adaptation, and, in particular, recommender systems. A random‐forests‐based modeling approach is shown as a generalization for earlier decision‐tree classification approach. Results of a usability survey conducted within the study are presented.