In order to understand the properties of online shopping that contribute to visitors’ purchasing habits, we have developed a Bayesian network model that infers the probability of purchase from eye movement and web log data. The results obtained from this model imply that a short visit time on catalog pages and a high frequency of fixation on all pages are related to increased purchase probability. Furthermore, it is shown that websites conforming to Internet Usability Guidelines (IUG)make visitors feel little stress regardless of browsing patterns, and that websites not conforming to IUG require a very short visit time on catalog pages if low stress is to be maintained.