Auctions have been a popular way of transaction on the Internet. Most of the studies of auction assume participants attending the auction are homogeneous. However, this assumption is open to question. In fact, every participant has his own personality, risk attitude, behavior, and cost when attending online auctions. This study takes an empirical approach and uses four variables, time of entry, time of exit, number of bids, and number of jump bids, to find the heterogeneity among bidders. We first used k-means clustering method to identify the types of bidders of online auctions, and then used C5.0 decision tree learning algorithm to find the rules to differentiate bidders. A taxonomy of four types of bidders is proposed in the study, which include observers, adventurers, opportunists, and early players. The results also suggest pacing of the auctions is an important factor that will affect bidder’s behavior in online auctions.