We study incremental clustering of objects that grow and accumulate over time. The objects come from a multi-table stream e.g. streams of
. As the Transactions stream accumulates, the Customers’ profiles
. First, we use an incremental propositionalisation to convert the multi-table stream into a single-table stream upon which we apply clustering. For this purpose, we develop an online version of K-Means algorithm that can handle these swelling objects and any new objects that arrive. The algorithm also
the quality of the model and performs re-clustering when it deteriorates. We evaluate our method on the PKDD Challenge 1999 dataset.