def hclust(pts, distance, clust_method="single", use_great_circle=True):
""" Hierarchical clustering method. Exposes R hclust function in Python.
Algorithm:
- Calculate distances for each pair of input pts
(use_great_circle: TRUE then use great_circle else use Euclidean)
- Cluster the points via hclust with distances and clust_method
- Cut the resultant clusters via cutree at the desired distance
- distance should be in kilometers IF use_great_circle=TRUE, otherwise,
they should be in the same metric as the input pts.
(see R doc for sp.spDist)
- Return the set of sets of points (i.e. the clusters)
"""
rpy2.robjects.numpy2ri.activate()
pt_array = numpy.array(pts)
r.library('sp')
sp_dists = r.spDists(pt_array, longlat=use_great_circle)
dists = r('as.dist')(sp_dists)
tree = r.hclust(dists, method=clust_method)
clusters = r.cutree(tree, h=distance)
# this is a little tricky
# clusters maintains a list of indexes for each point in the original list.
# The indexes represent the cluster number
# For example:
# clusters = [1, 3, 3, 1, 2, 1]
#
# where the 0th, 3rd, and last point in the original set
# belong to cluster number 1
#
# We want to return a list of clusters, each containing
# the index to the original point, so we map them here.
#
# Things are a little more confusing since R counts arrays
# from 1, python from 0 (hence the "- 1" from the cluster index)
list_of_pts = [[] for i in range(max(clusters))]
for j in range(0, len(clusters)):
list_of_pts[clusters[j] - 1].append(j)
return list_of_pts
#activate r to numpy array auto-conversion
rpy2.robjects.numpy2ri.activate()
#load sp library
r.library('sp')
to_x = numpy.random.rand(100) * 100
to_y = numpy.random.rand(100) * 100
xy = zip(to_x, to_y)
xy_array = numpy.array(xy)
dists = r.spDists(xy_array)
tree = r.hclust(r('as.dist')(dists), "average")
clusts = r.cutree(tree, h=100.0)
np_clusts = numpy.array(clusts)
clusters = [[] for i in range(max(np_clusts))]
for i in range(0, len(np_clusts)):
clusters[np_clusts[i] - 1].append(xy_array[i])
cluster1 = clusters[len(clusters) - 1]
cluster1_cols = numpy.array(cluster1)
print cluster1_cols
print cluster1[0]
dists = r.spDistsN1(cluster1_cols, cluster1[0])
print dists
rownames=['true'],
colnames=['predicted'])
# <headingcell level=4>
# Bridging the gap with Rpy2
# <codecell>
from rpy2.robjects import r
from rpy2.robjects.numpy2ri import numpy2ri as np2r
Xr = np2r(iris[['PW', 'PL', 'SW']].values)
d = r.dist(Xr)
tree = r.hclust(d, method='ward')
yhat_hclust = r.cutree(tree, k=3)
print pd.crosstab(iris['Type'],
yhat_hclust,
rownames=['true'],
colnames=['predicted'])
# <headingcell level=4>
# Using non-base packages in Rpy2
# <codecell>
import rpy2.robjects as robjects
from rpy2.robjects.packages import importr
r = robjects.r
yhat_new = new_label[yhat]
print pd.crosstab(iris['Type'], yhat_new, rownames=['true'], colnames=['predicted'])
# <headingcell level=4>
# Bridging the gap with Rpy2
# <codecell>
from rpy2.robjects import r
from rpy2.robjects.numpy2ri import numpy2ri as np2r
Xr = np2r(iris[['PW', 'PL', 'SW']].values)
d = r.dist(Xr)
tree = r.hclust(d, method='ward')
yhat_hclust = r.cutree(tree, k=3)
print pd.crosstab(iris['Type'], yhat_hclust, rownames=['true'], colnames=['predicted'])
# <headingcell level=4>
# Using non-base packages in Rpy2
# <codecell>
import rpy2.robjects as robjects
from rpy2.robjects.packages import importr
r = robjects.r
e1071 = importr('e1071')
Yr = np2r(iris['Type'])