Explain what method can be used to determine an optimal number of data clusters in cases when this number is unknown.