Python Programming for Biology: Bioinformatics and Beyond

Figure 23.3.  Example output of simple associative clustering at two thresholds. A

simple associative clustering method works by grouping points together into a cluster if

the separation between them (from one point to any other, considering all axes) is below a

specified threshold. Assuming the data is well separated and an appropriate threshold is

chosen, this method can be useful and is very fast. However, simple associative clustering

will naturally not separate overlapping clusters or situations where there are outliers,

within the clustering threshold, that can bridge between what otherwise appear to be

distinct groups.

The dbScanCluster function resembles simpleCluster but we add checks for the number

of neighbours and accordingly add any disconnected points to a list of noise, rather than a

cluster.  The  inputs  now  include  minNeighbour,  which  as  the  name  suggests  says  how

many neighbours an item should have at minimum.

def dbScanCluster(data, threshold, minNeighbour, distFunc=euclideanDist):

As before, the dictionary of neighbours is calculated and we initialise some variables,

including a set to hold the indices of the noise items.

neighbourDict = findNeighbours(data, distFunc, threshold)

clusters = []

noise = set()

pool = set(range(len(data)))

---