GA-based item partition for data mining

When a mining procedure is directly executed on very large databases, the computer memory may not allow the processing in memory. In the past, we adopted a branch-and-bound search strategy to divide the domain items as a set of groups. Although it works well in partitions the items, the time is quite time consuming. In this paper, we thus propose a GA-based approach to speed up the partition process. A new encoding representation and a transformation scheme are designed to help the search process. Experimental results also show that the algorithm can get a proper partition with good efficiency.