Approximate Query Answering in High-Dimensional Data Cubes