Mining evolving association rules for e-business recommendation

Association analysis is an effective data mining approach capable of unveiling interesting associations within a large dataset. Although widely adopted in e-business areas, it still has many difficulties when applied in practice. For instance, there is a mismatch between the static rules discovered and the drifting nature of the user interests, and it is difficult to detect associations from a huge volume of raw user data. This paper presents an effective approach to mine evolving association rules in order to tackle these problems. It is followed by a recommendation model based on the evolving association rules unveiled. Experimental results on an online toggery show that it can effectively unveil people’s shifting interests and make better recommendations accordingly.