While observing the user’s behavior we can make some assumptions about his preferences for BF or DF navigation. In a concept hierarchy we can associate a “level” (attribute) with each concept. The root concept has level 0. Its children have level 1, their children have level 2 and so on. We can compare the number of concepts visited at different levels.
In the strategy we can specify actions which are performed when the user accesses a concept. For example, if we see that the user visited more concepts at a higher level we increase the confidence that the user is prefers breadth-first structure. Otherwise we increase the confidence that he prefers depth-first structure. The author can specify this value for his application. After a number of steps when a threshold specified by the author is reached (30 or 70) we get a meaningful value of the attribute indicating the user’s preference.