To understand the trip patterns of subway passengers is very important to making plans for an efficient subway system. Accordingly, there have been studies on mining and classifying useful patterns from large smart card transaction databases of the Metropolitan Seoul subway system. In this paper, we define a new classification of subway trip patterns and devise a classification algorithm for eleven trip patterns of the subway users from smart card transaction databases which have been produced about ten million transactions daily. We have implemented the algorithm and then applied it to one-day transaction database to classify the trip patterns of subway passengers. We have focused on the analysis of significant patterns such as round-trip patterns, commuter patterns, and unexpected interesting patterns. The distribution of the number of passengers in each trip pattern is plotted by the get-on time and get-off time of subway transactions, which illustrates the characteristics of the significant patterns.