Abstract
In the past, bus service planning in metropolitan area was a crucial procedure of bus operators, private organizations, or local governments. One of the important topics of bus service improvement is of course how to understand the actual users' decisions, or strong points with which they can attract the users to their bus, in other words. However, there are only a small minority of users announce their thoughts directly. Therefore, in order to understand user's likes or dislikes, we use a time series data analysis technique to large e-ticket data. In order to obtain user's decisions derived from e-ticket system, we propose this new method to cluster the users via user behavior. User behavior transition map is to be drawn from continuous behavior transition data. A quantitative evaluation of user's decisions will be shown in the transition result, and therefore can target cluster by cluster to realize each need.
