Recently, a new media service such as IPTV has started so that tremendous amounts of TV program contents become available at the user sides. Also, users can access many TV program contents via multi-channels. Therefore, research on automatic TV program recommendation has been made to provide easy access to the TV programs for users to alleviate the user``s burden in finding their preferred TV program contents. On the other hand, the TV watching behaviors of users exhibit somewhat regular TV watching patterns specific to particular users in chronological and/or semantic perspectives. This means that the watched TV program contents are semantically or chronologically related. This phenomenon can be easily found in goods purchases. That is, the purchased items are related sequentially in a period of time.
In this thesis, automatic and personalized TV program schedule recommendation is studied so as to make it easier to watch (IP)TV program contents sequentially. It is believed based upon our best knowledge that the concept of automatic and personalized TV program scheduler recommendation be first studied. In this thesis, the TV program scheduler recommendation is made by using sequential pattern mining in conjunction with collaborative filtering.
The prefix span algorithm utilizes the number of purchased items to construct the sequences of frequently purchased items. In this thesis, we extend the prefix span algorithm by incorporating the user``s interest in TV program contents, the total time lengths and the watched time lengths of TV program contents into the construction of recommended sequences of TV programs contents. Furthermore, the recommended sequences of frequently watched TV program contents are refined by taking into account the user preference on program content itself, genre and channel. To validate our proposed scheme, we use the TV watching history of 30 users for six months in which the data of the first three months are used for training and ...