Merge Center Clustering (MCC)

Merge center clustering (MCC) is a sweet alternative to the omnipresent transitive clustering. Both algorithms are low in computational cost, which is necessary in most entity resolution scenarios. However, according to several scientific studies, MCC is superior to transitive clustering regarding resolution quality.

We will demonstrate its workings with a toy dataset of eleven records from the following code cell. Every line represents a match prediction from the pairwise model. An absent combination means it did not make it through the indexing step, or the pairwise model predicts a no-match for that pair.