Usage of dynamic web applications to solve office or personal tasks has become part of our everyday life. Way how people look, while solving these tasks, reflects not only what task are they solving, but also their level in given domain or personal characteristics. Scanpath is formed by fixations and saccades in time and models, which use only statistical informations about scanpaths are losing important information. Probabilistic methods present a new way, how to model eye movement as time-series information and can efficiently capture stochastic character of scanpath.
In this work I am using probabilistic models like Hidden Markov Model or Gaussian models to create features and model eye movement data. Based on used data and selected domain, I will be predicting task at hand or user characteristics.