I know the correct answer is DBSCAN clustering using the Box-Graph, i just don't know why exactly. Why don't the other works?

Which one of the following the algorithms from the previous lectures still work if we replace the Euclidean space with the metric space defined on input point set? If the algorithm does not work anymore, which assumption did the algorithm make that does not hold any longer? (a) k-means clustering using Lloyd algorithm initilialized with k random points from the input (b) DBSCAN clustering using the graph-approach (DBSCAN1) (c) DBSCAN clustering using the Box-Graph (d) single-link agglomerative clustering Which one of the following the algorithms from the previous lectures still work if we replace the Euclidean space with the metric space defined on input point set? If the algorithm does not work anymore, which assumption did the algorithm make that does not hold any longer? (a) k-means clustering using Lloyd algorithm initilialized with k random points from the input (b) DBSCAN clustering using the graph-approach (DBSCAN1) (c) DBSCAN clustering using the Box-Graph (d) single-link agglomerative clustering