Detecting users’ usage intentions for websites employing deep learning on eye-tracking data

We proposed a method employing deep learning (DL) on eye-tracking data and applied this method to detect intentions to use apparel websites that differed in factors of depth, breadth, and location of navigation. Results showed that users’ intentions could be predicted by combining a deep neural network algorithm and metrics recorded from an eye-tracker. Using all of the eye-tracking metric features attained the best accuracy when predicting usage/not-usage intention to websites. In addition, the results suggest that for apparel websites with the same depth, designers can increase usage intention by using a larger number of navigation items and placing the navigation at the top and left of the homepage. The results show that building intelligent usage intention-detection systems is possible for the range of websites we examined and is also computationally practical. Hence, the study motivates future investigations that focus on design of such systems.