Hard and soft updating centroids for clustering Y-short tandem repeats (Y-STR) data
Ali Seman, Zahari Abu Bakar, Nuraina Daud · 2010
This paper compares hard and soft updating centroids for clustering Y-STR data. The hard centroids represented by New Fuzzy k-Modes clustering algorithm, whereas the soft centroids represented through k-Population algorithm. These two algorithms are experimented through two datasets, Y-STR haplogroups and Y-STR Surnames. The results show that the soft centroid performance is better than the hard centroid for Y-STR data. The soft centroid produces 86.3% of the average clustering accuracy as compared 84.3% of the new fuzzy k-Modes algorithm. However, the overall result shows that the hard updating clustering is better than the soft updating clustering while clustering Y-STR data.