, there are promising alternatives available. For instance constraint K-Means avoids creating empty clusters the way K-means does, by introducing an extra step defining constraints for the algorithm at the beginning [20]. Also weighted K-Means could be used as it calculates each cluster’s centroid not by simply calculating the mean of all attribute values but by assigning weights to these [11]. This way each metric can be assigned weights using expert opinion. Finally we can avoid having to define the number of clusters in advance, and to depend on initial centre definition, if we employ clustering algorithms automating this process [7].