A&C_IADI Logo
HOME ABOUT VOLUMES SEARCH

Volumes / Journal / 27 - 2016 / Collins-Elliott S.

Agglomerative clustering using cosine and Jaccard distances: a computational approach to Roman vessel taxonomy

Click on any image to open slideshow

Fig. 1 – Scatter plot of vessel types using the second and third axes resulting from singular value decomposition of the term-document matrix generated from the pilot dataset, using only vessel ware attributes, with and without labels for each record.

Fig. 1 – Scatter plot of vessel types using the second and third axes resulting ...

Fig. 2 – Dendrogram of the six vectors from Table 3 using Ward’s method. The cosine distance matrix revealed identical groupings as the Jaccard distance.

Fig. 2 – Dendrogram of the six vectors from Table 3 using Ward’s method. The cos...

Fig. 3 – Dendrograms of agglomerative clusters using Ward’s method (truncated), with a cut-off distance of 10. Vessel-type clusters in the domain of vessel form (above) and ware (below). Dendrograms on the left were based on cosine distances, those on the right on Jaccard distances.

Fig. 3 – Dendrograms of agglomerative clusters using Ward’s method (truncated), ...

Fig. 4 – Dendrograms of agglomerative clusters using Ward’s method (truncated), with a cut-off distance of 50, of vessel-type clusters based on form. Dendrogram on the left was based on cosine distances, that on the right on Jaccard distances.

Fig. 4 – Dendrograms of agglomerative clusters using Ward’s method (truncated), ...

Collins-Elliott S. 2016, Agglomerative clustering using cosine and Jaccard distances: a computational approach to Roman vessel taxonomy, Archeologia e Calcolatori 27, 10.19282/AC.27.2016.01