def test_dataset():
"""Test the Dataset class."""
print ' Testing class Dataset'
# TEST CASE 1
# Create and test an empty dataset
dset1 = a6.Dataset(3)
cornelltest.assert_equals(3, dset1.getDimension())
cornelltest.assert_equals(0, dset1.getSize())
# We use this assert function to compare lists
cornelltest.assert_float_lists_equal([], dset1.getContents())
print ' Default initialization looks okay'
# TEST CASE 2
# Create and test a non-empty dataset
items = [[0.0, 0.0, 0.0], [1.0, 0.0, 0.0], [0.0, 1.0, 0.0],
[0.0, 0.0, 1.0]]
dset2 = a6.Dataset(3, items)
cornelltest.assert_equals(3, dset2.getDimension())
cornelltest.assert_equals(4, dset2.getSize())
# Check that contents is initialized correctly
# Make sure items is COPIED
cornelltest.assert_float_lists_equal(items, dset2.getContents())
cornelltest.assert_false(dset2.getContents() is items)
cornelltest.assert_false(dset2.getContents()[0] is items[0])
print ' User-provided initialization looks okay'
# Check that getPoint() is correct AND that it copies
cornelltest.assert_float_lists_equal([0.0, 1.0, 0.0], dset2.getPoint(2))
cornelltest.assert_false(dset2.getContents()[2] is dset2.getPoint(2))
print ' Method Dataset.getPoint looks okay'
# Add something to the dataset (and check it was added)
dset1.addPoint([0.0, 0.5, 4.2])
cornelltest.assert_float_lists_equal([[0.0, 0.5, 4.2]],
dset1.getContents())
cornelltest.assert_float_lists_equal([0.0, 0.5, 4.2], dset1.getPoint(0))
# Check the point is COPIED
cornelltest.assert_false(dset1.getPoint(0) is dset1.getContents()[0])
extra = [0.0, 0.5, 4.2]
dset2.addPoint(extra)
items.append(extra)
cornelltest.assert_float_lists_equal(items, dset2.getContents())
# Check the point was COPIED
cornelltest.assert_false(id(extra) in map(id, dset2.getContents()))
print ' Method Dataset.addPoint looks okay'
print ' class Dataset appears correct'
print ''
def _load_file(self, filename):
"""Load a data set file into a Dataset."""
try:
f = open(filename)
contents = []
first = f.readline()
assert first[1:].strip()[:5] == 'kdata'
for x in f:
if x[0] != '#': # Ignore comments
point = parse_data(x)
contents.append(point)
self._ds = a6.Dataset(3, contents)
if not self._ds.getContents():
raise RuntimeError()
shortname = os.path.split(filename)[1]
if (len(shortname) > 10):
shortname = shortname[0:10] + '...'
self._filebutton.configure(text=shortname)
self._kmean = None
self._plot()
except RuntimeError:
tkMessageBox.showwarning(
'Load', 'ERROR: You must complete Dataset first.')
except AssertionError:
tkMessageBox.showwarning('Load',
'ERROR: CSV file must be a kdata file.')
except:
traceback.print_exc()
tkMessageBox.showwarning('Load', 'ERROR: kdata file is corrupted.')
def test_kmeans_a():
"""Test Part A of the ClusterGroup class."""
print ' Testing Part A of class ClusterGroup'
# A dataset with four points almost in a square
items = [[0.,0.], [10.,1.], [10.,10.], [0.,9.]]
dset = a6.Dataset(2, items)
# Test creating a clustering with random seeds
km = a6.ClusterGroup(dset, 3)
# Should have 3 clusters
cornelltest.assert_equals(len(km.getClusters()), 3)
for clust in km.getClusters():
# cluster centroids should have been chosen from items
cornelltest.assert_true(clust.getCentroid() in items)
# cluster centroids should be distinct (since items are)
for clust2 in km.getClusters():
if clust2 is not clust:
cornelltest.assert_float_lists_not_equal(clust.getCentroid(), clust2.getCentroid())
print ' Random ClusterGroup initialization looks okay'
# Clusterings of that dataset, with two and three deterministic clusters
km = a6.ClusterGroup(dset, 2, [0,2])
cornelltest.assert_equals(items[0], km.getClusters()[0].getCentroid())
cornelltest.assert_equals(items[2], km.getClusters()[1].getCentroid())
km = a6.ClusterGroup(dset, 3, [0,2,3])
cornelltest.assert_equals(items[0], km.getClusters()[0].getCentroid())
cornelltest.assert_equals(items[2], km.getClusters()[1].getCentroid())
cornelltest.assert_equals(items[3], km.getClusters()[2].getCentroid())
print ' Seeded ClusterGroup initialization looks okay'
print ' Part A of class ClusterGroup appears correct'
print ''
def test_kmeans_b():
"""Test Part B of the ClusterGroup class."""
# This function tests the methods _nearest_cluster and _partition,
# both of which are private methods. Normally it's not good form to
# directly call these methods from outside the class, but we make an
# exception for testing code, which often has to be more tightly
# integrated with the implementation of a class than other code that
# just uses the class.
print ' Testing Part B of class ClusterGroup'
# Reinitialize data set
items = [[0., 0.], [10., 1.], [10., 10.], [0., 9.]]
dset = a6.Dataset(2, items)
km1 = a6.ClusterGroup(dset, 2, [0, 2])
km2 = a6.ClusterGroup(dset, 3, [0, 2, 3])
nearest = km1._nearest_cluster([1., 1.])
cornelltest.assert_true(nearest is km1.getClusters()[0])
nearest = km1._nearest_cluster([1., 10.])
cornelltest.assert_true(nearest is km1.getClusters()[1])
nearest = km2._nearest_cluster([1., 1.])
cornelltest.assert_true(nearest is km2.getClusters()[0])
nearest = km2._nearest_cluster([1., 10.])
cornelltest.assert_true(nearest is km2.getClusters()[2])
print ' Method ClusterGroup._nearest_cluster() looks okay'
# Testing partition()
# For this example points 0 and 3 are closer, as are 1 and 2
km1._partition()
cornelltest.assert_equals(set([0, 3]),
set(km1.getClusters()[0].getIndices()))
cornelltest.assert_equals(set([1, 2]),
set(km1.getClusters()[1].getIndices()))
# partition and repeat -- should not change clusters.
km1._partition()
cornelltest.assert_equals(set([0, 3]),
set(km1.getClusters()[0].getIndices()))
cornelltest.assert_equals(set([1, 2]),
set(km1.getClusters()[1].getIndices()))
# Reset the cluster centroids; now it changes
cluster = km1.getClusters()
cluster[0]._centroid = [5.0, 10.0]
cluster[1]._centroid = [0.0, 2.0]
km1._partition()
cornelltest.assert_equals(set([2, 3]),
set(km1.getClusters()[0].getIndices()))
cornelltest.assert_equals(set([0, 1]),
set(km1.getClusters()[1].getIndices()))
print ' Method ClusterGroup._partition() looks okay'
print ' Part B of class ClusterGroup appears correct'
print ''
def test_cluster_b():
"""Test Part B of the Cluster class assignment."""
print ' Testing Part B of class Cluster'
# A dataset with four points
items = [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 0.0],
[0.0, 0.0, 1.0]]
dset = a6.Dataset(3, items)
# Create two clusters
cluster2 = a6.Cluster(dset, [0.5, 0.5, 0.0])
cluster3 = a6.Cluster(dset, [0.0, 0.0, 0.5])
# TEST CASE 1 (distance)
dist = cluster2.distance([1.0, 0.0, -1.0])
cornelltest.assert_floats_equal(1.22474487139, dist)
# TEST CASE 2 (distance)
dist = cluster2.distance([0.5, 0.5, 0.0])
cornelltest.assert_floats_equal(0.0, dist)
# TEST CASE 3 (distance)
dist = cluster3.distance([0.5, 0.0, 0.5])
cornelltest.assert_floats_equal(0.5, dist)
print ' Method Cluster.distance() looks okay'
# TEST CASE 1 (updateCentroid): centroid remains the same
cluster2.addIndex(0)
cluster2.addIndex(1)
stable = cluster2.updateCentroid()
cornelltest.assert_float_lists_equal([0.5, 0.5, 0.0],
cluster2.getCentroid())
cornelltest.assert_true(stable)
# TEST CASE 2 (updateCentroid): centroid changes
cluster2.addIndex(2)
cluster2.addIndex(3)
stable = cluster2.updateCentroid()
cornelltest.assert_float_lists_equal([0.25, 0.25, 0.25],
cluster2.getCentroid())
cornelltest.assert_false(stable)
# updating again without changing points: centroid stable
stable = cluster2.updateCentroid()
cornelltest.assert_float_lists_equal([0.25, 0.25, 0.25],
cluster2.getCentroid())
cornelltest.assert_true(stable)
print ' Method Cluster.updateCentroid() looks okay'
print ' Part B of class Cluster appears correct'
print ''
def test_cluster_a():
"""Test Part A of the Cluster class assignment."""
print ' Testing Part A of class Cluster'
# TEST CASE 1
# Create and test a cluster (always empty)
dset = a6.Dataset(3)
point = [0.0,1.0,0.0]
cluster1 = a6.Cluster(dset, point)
# Compare centroid and contents
cornelltest.assert_float_lists_equal(point,cluster1.getCentroid())
cornelltest.assert_equals([],cluster1.getIndices())
# Make sure centroid COPIED
cornelltest.assert_not_equals(id(point),id(cluster1.getCentroid()))
print ' Basic cluster methods look okay'
# Add something to cluster (and check it was added)
extra = [[0.0,0.5,4.2],[0.0,1.0,0.0]]
dset.addPoint(extra[0])
dset.addPoint(extra[1])
cluster1.addIndex(1)
cornelltest.assert_equals([1],cluster1.getIndices())
cluster1.addIndex(0)
cornelltest.assert_equals([1,0],cluster1.getIndices())
# Make sure we can handle duplicates!
cluster1.addIndex(1)
cornelltest.assert_equals([1,0],cluster1.getIndices())
print ' Method Cluster.addIndex look okay'
# And clear it
contents = cluster1.getContents()
cornelltest.assert_equals(2,len(contents))
cornelltest.assert_float_lists_equal(extra[1],contents[0])
cornelltest.assert_float_lists_equal(extra[0],contents[1])
print ' Method Cluster.getContents look okay'
# And clear it
cluster1.clear()
cornelltest.assert_equals([],cluster1.getIndices())
print ' Method Cluster.clear look okay'
print ' Part A of class Cluster appears correct'
print ''
def candy_to_kmeans(filename,k,seed=None):
"""Returns: KMeans object for the given candy CSV file.
Candy CSV files have 5 attributes: the candy name, the sweetness, the sourness,
the nuttiness, and the texture. The first value is a string. The remaining
four values are floats between 0 and 1.
Precondition: filename is a name of a candy CSV file."""
file = open(filename)
contents = []
for x in file:
if x[0] != '#': # Ignore comments
point = x.strip().split(',')
point = map(float,point[1:]) # Remove the name and convert to floats
contents.append(point)
dataset = a6.Dataset(4,contents)
return a6.ClusterGroup(dataset, k, seed)
def test_kmeans_d():
"""Test Part D of the ClusterGroup class."""
print ' Testing Part D of class ClusterGroup'
items = [[0.5,0.5,0.5],[0.5,0.6,0.6],[0.6,0.5,0.6],[0.5,0.6,0.5],[0.5,0.4,0.5],[0.5,0.4,0.4]]
dset = a6.Dataset(3,items)
# Try the same test case straight from the top using perform_k_means
km1 = a6.ClusterGroup(dset, 2, [1, 3])
km1.run(10)
# Check first cluster
cluster1 = km1.getClusters()[0]
cornelltest.assert_float_lists_equal([8./15, 17./30, 17./30], cluster1.getCentroid())
cornelltest.assert_equals(set([1, 2, 3]), set(cluster1.getIndices()))
# Check second cluster
cluster2 = km1.getClusters()[1]
cornelltest.assert_float_lists_equal([0.5, 13./30, 14./30],cluster2.getCentroid())
cornelltest.assert_equals(set([0, 4, 5]), set(cluster2.getIndices()))
print ' Method run looks okay'
# Test on a real world data set
km2 = candy_to_kmeans('datasets/small_candy.csv',3,[23, 54, 36])
km2.run(20)
# The actual results
cluster0 = km2.getClusters()[0]
cluster1 = km2.getClusters()[1]
cluster2 = km2.getClusters()[2]
# The "correct" answers
contents0 = [[0.88, 0.84, 0.8, 0.3], [0.02, 0.67, 0.75, 0.61], [0.81, 0.69, 0.65, 0.65],
[0.62, 0.75, 0.65, 0.43], [0.35, 0.63, 0.65, 0.12], [0.61, 0.85, 0.81, 0.44],
[0.95, 0.94, 0.98, 0.69], [0.04, 0.69, 0.38, 0.39], [0.28, 0.91, 0.63, 0.08],
[0.38, 0.94, 0.53, 0.07], [0.08, 0.62, 0.32, 0.27], [0.69, 0.82, 0.75, 0.65],
[0.84, 0.89, 0.91, 0.38], [0.22, 0.88, 0.39, 0.33], [0.71, 0.78, 0.64, 0.57],
[0.15, 0.87, 0.62, 0.22], [0.65, 0.81, 0.69, 0.55], [0.27, 0.63, 0.69, 0.39],
[0.35, 0.7, 0.41, 0.15], [0.91, 0.98, 0.61, 0.58], [0.9, 0.63, 0.83, 0.6],
[0.95, 0.83, 0.64, 0.5], [0.76, 0.86, 0.74, 0.61], [0.27, 0.65, 0.52, 0.28],
[0.86, 0.91, 0.88, 0.62], [0.1, 0.79, 0.5, 0.12], [0.99, 0.68, 0.8, 0.42],
[0.09, 0.85, 0.55, 0.21], [0.79, 0.94, 0.83, 0.48], [0.73, 0.92, 0.74, 0.39],
[0.89, 0.72, 0.78, 0.38], [0.39, 0.9, 0.52, 0.26], [0.46, 0.35, 0.96, 0.05],
[0.21, 0.62, 0.33, 0.09], [0.58, 0.37, 0.9, 0.08], [0.54, 0.92, 0.36, 0.35],
[0.67, 0.32, 0.66, 0.2], [0.36, 0.64, 0.57, 0.26], [0.9, 0.7, 0.74, 0.63],
[0.4, 0.69, 0.74, 0.7]]
contents1 = [[0.32, 0.87, 0.14, 0.68], [0.73, 0.31, 0.15, 0.08], [0.87, 0.99, 0.2, 0.8],
[0.77, 0.45, 0.31, 0.31], [0.96, 0.09, 0.49, 0.3], [0.86, 0.03, 0.3, 0.39],
[0.86, 0.86, 0.32, 0.88], [0.8, 0.4, 0.23, 0.33], [0.81, 0.66, 0.26, 0.82],
[0.95, 0.62, 0.28, 0.01], [0.35, 0.71, 0.01, 0.32], [0.73, 0.65, 0.23, 0.02],
[0.84, 0.88, 0.04, 0.86], [0.8, 0.62, 0.09, 0.65], [0.72, 0.55, 0.1, 0.17],
[0.61, 0.42, 0.24, 0.33], [0.72, 0.88, 0.02, 0.95], [0.88, 0.96, 0.09, 0.88],
[0.9, 0.05, 0.34, 0.41], [0.9, 0.41, 0.27, 0.36]]
contents2 = [[0.4, 0.21, 0.78, 0.68], [0.54, 0.06, 0.81, 0.98], [0.2, 0.54, 0.73, 0.85],
[0.14, 0.31, 0.86, 0.74], [0.39, 0.14, 0.99, 0.24], [0.23, 0.32, 0.7, 0.75],
[0.65, 0.05, 0.39, 0.49], [0.04, 0.52, 0.99, 0.75], [0.14, 0.55, 0.67, 0.63],
[0.5, 0.2, 0.69, 0.95], [0.79, 0.09, 0.41, 0.69], [0.4, 0.3, 0.78, 0.74],
[0.65, 0.24, 0.63, 0.27], [0.35, 0.3, 0.94, 0.92], [0.39, 0.38, 0.85, 0.32],
[0.38, 0.07, 0.82, 0.01], [0.66, 0.09, 0.69, 0.46], [0.26, 0.39, 0.95, 0.63],
[0.54, 0.06, 0.74, 0.86], [0.2, 0.48, 0.98, 0.84], [0.62, 0.24, 0.77, 0.17],
[0.27, 0.38, 0.76, 0.63], [0.7, 0.04, 0.7, 0.82], [0.41, 0.11, 0.61, 0.78],
[0.22, 0.44, 0.67, 0.99], [0.51, 0.05, 0.95, 0.66], [0.44, 0.1, 0.61, 0.98],
[0.31, 0.16, 0.95, 0.9], [0.31, 0.5, 0.87, 0.85], [0.5, 0.09, 0.84, 0.78],
[0.62, 0.01, 0.88, 0.1], [0.44, 0.28, 0.88, 0.99], [0.57, 0.23, 0.6, 0.85],
[0.72, 0.14, 0.63, 0.37], [0.39, 0.08, 0.77, 0.96], [0.09, 0.47, 0.63, 0.8],
[0.63, 0.05, 0.52, 0.63], [0.62, 0.27, 0.67, 0.77], [0.35, 0.04, 0.85, 0.86],
[0.36, 0.34, 0.75, 0.37]]
centroid0 = [0.54125, 0.7545, 0.66125, 0.3775]
centroid1 = [0.76900, 0.5705, 0.20550, 0.4775]
centroid2 = [0.42325, 0.2330, 0.75775, 0.6765]
cornelltest.assert_float_lists_equal(centroid0,cluster0.getCentroid())
cornelltest.assert_float_lists_equal(centroid1,cluster1.getCentroid())
cornelltest.assert_float_lists_equal(centroid2,cluster2.getCentroid())
cornelltest.assert_float_lists_equal(contents0,cluster0.getContents())
cornelltest.assert_float_lists_equal(contents1,cluster1.getContents())
cornelltest.assert_float_lists_equal(contents2,cluster2.getContents())
print ' Candy analysis test looks okay'
print ' Part D of class ClusterGroup appears correct'
print ''
def test_kmeans_c():
"""Test Part C of the ClusterGroup class."""
print ' Testing Part C of class ClusterGroup'
items = [[0.,0.], [10.,1.], [10.,10.], [0.,9.]]
dset = a6.Dataset(2, items)
km1 = a6.ClusterGroup(dset, 2, [0,2])
km1._partition()
# Test update()
stable = km1._update()
cornelltest.assert_float_lists_equal([0,4.5], km1.getClusters()[0].getCentroid())
cornelltest.assert_float_lists_equal([10.0,5.5], km1.getClusters()[1].getCentroid())
cornelltest.assert_false(stable)
# updating again should not change anything, but should return stable
stable = km1._update()
cornelltest.assert_float_lists_equal([0,4.5], km1.getClusters()[0].getCentroid())
cornelltest.assert_float_lists_equal([10.0,5.5], km1.getClusters()[1].getCentroid())
cornelltest.assert_true(stable)
print ' Method ClusterGroup._update() looks okay'
# Now test the k-means process itself.
# FOR ALL TEST CASES
# Create and initialize a non-empty dataset
items = [[0.5,0.5,0.5],[0.5,0.6,0.6],[0.6,0.5,0.6],[0.5,0.6,0.5],[0.5,0.4,0.5],[0.5,0.4,0.4]]
dset = a6.Dataset(3,items)
# Create a clustering, providing non-random seed indices so the test is deterministic
km2 = a6.ClusterGroup(dset, 2, [1, 3])
# PRE-TEST: Check first cluster (should be okay if passed part D)
cluster1 = km2.getClusters()[0]
cornelltest.assert_float_lists_equal([0.5, 0.6, 0.6], cluster1.getCentroid())
cornelltest.assert_equals(set([]), set(cluster1.getIndices()))
# PRE-TEST: Check second cluster (should be okay if passed part D)
cluster2 = km2.getClusters()[1]
cornelltest.assert_float_lists_equal([0.5, 0.6, 0.5], cluster2.getCentroid())
cornelltest.assert_equals(set([]), set(cluster2.getIndices()))
# Make a fake cluster to test update_centroid() method
clustertest = a6.Cluster(dset, [0.5, 0.6, 0.6])
for ind in [1, 2]:
clustertest.addIndex(ind)
# TEST CASE 1 (update)
stable = clustertest.updateCentroid()
cornelltest.assert_float_lists_equal([0.55, 0.55, 0.6],clustertest.getCentroid())
cornelltest.assert_false(stable) # Not yet stable
# TEST CASE 2 (update)
stable = clustertest.updateCentroid()
cornelltest.assert_float_lists_equal([0.55, 0.55, 0.6],clustertest.getCentroid())
cornelltest.assert_true(stable) # Now it is stable
# TEST CASE 3 (step)
km2.step()
# Check first cluster (WHICH HAS CHANGED!)
cluster1 = km2.getClusters()[0]
cornelltest.assert_float_lists_equal([0.55, 0.55, 0.6], cluster1.getCentroid())
cornelltest.assert_equals(set([1, 2]), set(cluster1.getIndices()))
# Check second cluster (WHICH HAS CHANGED!)
cluster2 = km2.getClusters()[1]
cornelltest.assert_float_lists_equal([0.5, 0.475, 0.475],cluster2.getCentroid())
cornelltest.assert_equals(set([0, 3, 4, 5]), set(cluster2.getIndices()))
# TEST CASE 3 (step)
km2.step()
# Check first cluster (WHICH HAS CHANGED!)
cluster1 = km2.getClusters()[0]
cornelltest.assert_float_lists_equal([8./15, 17./30, 17./30], cluster1.getCentroid())
cornelltest.assert_equals(set([1, 2, 3]), set(cluster1.getIndices()))
# Check second cluster (WHICH HAS CHANGED!)
cluster2 = km2.getClusters()[1]
cornelltest.assert_float_lists_equal([0.5, 13./30, 14./30],cluster2.getCentroid())
cornelltest.assert_equals(set([0, 4, 5]), set(cluster2.getIndices()))
print ' Method ClusterGroup.step looks okay'
print ' Part C of class ClusterGroup appears correct'
print ''
def test_kmeans_c():
"""Test Part C of the ClusterGroup class."""
print ' Testing Part C of class ClusterGroup'
items = [[0., 0.], [10., 1.], [10., 10.], [0., 9.]]
dset = a6.Dataset(2, items)
km1 = a6.ClusterGroup(dset, 2, [0, 2])
km1._partition()
# Test update()
stable = km1._update()
cornelltest.assert_float_lists_equal([0, 4.5],
km1.getClusters()[0].getCentroid())
cornelltest.assert_float_lists_equal([10.0, 5.5],
km1.getClusters()[1].getCentroid())
cornelltest.assert_false(stable)
# updating again should not change anything, but should return stable
stable = km1._update()
cornelltest.assert_float_lists_equal([0, 4.5],
km1.getClusters()[0].getCentroid())
cornelltest.assert_float_lists_equal([10.0, 5.5],
km1.getClusters()[1].getCentroid())
cornelltest.assert_true(stable)
print ' Method ClusterGroup._update() looks okay'
# Now test the k-means process itself.
# FOR ALL TEST CASES
# Create and initialize a non-empty dataset
items = [[0.5, 0.5, 0.5], [0.5, 0.6, 0.6], [0.6, 0.5, 0.6],
[0.5, 0.6, 0.5], [0.5, 0.4, 0.5], [0.5, 0.4, 0.4]]
dset = a6.Dataset(3, items)
# Create a clustering, providing non-random seed indices so the test is deterministic
km2 = a6.ClusterGroup(dset, 2, [1, 3])
# PRE-TEST: Check first cluster (should be okay if passed part D)
cluster1 = km2.getClusters()[0]
cornelltest.assert_float_lists_equal([0.5, 0.6, 0.6],
cluster1.getCentroid())
cornelltest.assert_equals(set([]), set(cluster1.getIndices()))
# PRE-TEST: Check second cluster (should be okay if passed part D)
cluster2 = km2.getClusters()[1]
cornelltest.assert_float_lists_equal([0.5, 0.6, 0.5],
cluster2.getCentroid())
cornelltest.assert_equals(set([]), set(cluster2.getIndices()))
# Make a fake cluster to test update_centroid() method
clustertest = a6.Cluster(dset, [0.5, 0.6, 0.6])
for ind in [1, 2]:
clustertest.addIndex(ind)
# TEST CASE 1 (update)
stable = clustertest.updateCentroid()
cornelltest.assert_float_lists_equal([0.55, 0.55, 0.6],
clustertest.getCentroid())
cornelltest.assert_false(stable) # Not yet stable
# TEST CASE 2 (update)
stable = clustertest.updateCentroid()
cornelltest.assert_float_lists_equal([0.55, 0.55, 0.6],
clustertest.getCentroid())
cornelltest.assert_true(stable) # Now it is stable
# TEST CASE 3 (step)
km2.step()
# Check first cluster (WHICH HAS CHANGED!)
cluster1 = km2.getClusters()[0]
cornelltest.assert_float_lists_equal([0.55, 0.55, 0.6],
cluster1.getCentroid())
cornelltest.assert_equals(set([1, 2]), set(cluster1.getIndices()))
# Check second cluster (WHICH HAS CHANGED!)
cluster2 = km2.getClusters()[1]
cornelltest.assert_float_lists_equal([0.5, 0.475, 0.475],
cluster2.getCentroid())
cornelltest.assert_equals(set([0, 3, 4, 5]), set(cluster2.getIndices()))
# TEST CASE 3 (step)
km2.step()
# Check first cluster (WHICH HAS CHANGED!)
cluster1 = km2.getClusters()[0]
cornelltest.assert_float_lists_equal([8. / 15, 17. / 30, 17. / 30],
cluster1.getCentroid())
cornelltest.assert_equals(set([1, 2, 3]), set(cluster1.getIndices()))
# Check second cluster (WHICH HAS CHANGED!)
cluster2 = km2.getClusters()[1]
cornelltest.assert_float_lists_equal([0.5, 13. / 30, 14. / 30],
cluster2.getCentroid())
cornelltest.assert_equals(set([0, 4, 5]), set(cluster2.getIndices()))
# Try it on a file
km3 = candy_to_kmeans('datasets/smallcandy.csv', 3, [23, 54, 36])
km3.step()
# The actual results
cluster0 = km3.getClusters()[0]
cluster1 = km3.getClusters()[1]
cluster2 = km3.getClusters()[2]
# The "correct" answers
contents0 = [[0.88, 0.84, 0.8, 0.3], [0.02, 0.67, 0.75, 0.61],
[0.2, 0.54, 0.73, 0.85], [0.62, 0.75, 0.65, 0.43],
[0.35, 0.63, 0.65, 0.12], [0.61, 0.85, 0.81, 0.44],
[0.95, 0.94, 0.98, 0.69], [0.04, 0.69, 0.38, 0.39],
[0.04, 0.52, 0.99, 0.75], [0.28, 0.91, 0.63, 0.08],
[0.14, 0.55, 0.67, 0.63], [0.38, 0.94, 0.53, 0.07],
[0.08, 0.62, 0.32, 0.27], [0.69, 0.82, 0.75, 0.65],
[0.84, 0.89, 0.91, 0.38], [0.22, 0.88, 0.39, 0.33],
[0.39, 0.38, 0.85, 0.32], [0.26, 0.39, 0.95, 0.63],
[0.15, 0.87, 0.62, 0.22], [0.65, 0.81, 0.69, 0.55],
[0.27, 0.63, 0.69, 0.39], [0.35, 0.7, 0.41, 0.15],
[0.2, 0.48, 0.98, 0.84], [0.76, 0.86, 0.74, 0.61],
[0.27, 0.65, 0.52, 0.28], [0.86, 0.91, 0.88, 0.62],
[0.1, 0.79, 0.5, 0.12], [0.09, 0.85, 0.55, 0.21],
[0.79, 0.94, 0.83, 0.48], [0.73, 0.92, 0.74, 0.39],
[0.31, 0.5, 0.87, 0.85], [0.39, 0.9, 0.52, 0.26],
[0.46, 0.35, 0.96, 0.05], [0.21, 0.62, 0.33, 0.09],
[0.58, 0.37, 0.9, 0.08], [0.54, 0.92, 0.36, 0.35],
[0.36, 0.64, 0.57, 0.26], [0.09, 0.47, 0.63, 0.8],
[0.4, 0.69, 0.74, 0.7]]
contents1 = [[0.32, 0.87, 0.14, 0.68], [0.87, 0.99, 0.2, 0.8],
[0.86, 0.86, 0.32, 0.88], [0.81, 0.66, 0.26, 0.82],
[0.91, 0.98, 0.61, 0.58], [0.84, 0.88, 0.04, 0.86],
[0.8, 0.62, 0.09, 0.65], [0.72, 0.88, 0.02, 0.95],
[0.88, 0.96, 0.09, 0.88]]
contents2 = [[0.4, 0.21, 0.78, 0.68], [0.54, 0.06, 0.81, 0.98],
[0.73, 0.31, 0.15, 0.08], [0.81, 0.69, 0.65, 0.65],
[0.14, 0.31, 0.86, 0.74], [0.77, 0.45, 0.31, 0.31],
[0.39, 0.14, 0.99, 0.24], [0.23, 0.32, 0.7, 0.75],
[0.65, 0.05, 0.39, 0.49], [0.96, 0.09, 0.49, 0.3],
[0.86, 0.03, 0.3, 0.39], [0.5, 0.2, 0.69, 0.95],
[0.79, 0.09, 0.41, 0.69], [0.4, 0.3, 0.78, 0.74],
[0.65, 0.24, 0.63, 0.27], [0.35, 0.3, 0.94, 0.92],
[0.71, 0.78, 0.64, 0.57], [0.8, 0.4, 0.23, 0.33],
[0.38, 0.07, 0.82, 0.01], [0.66, 0.09, 0.69, 0.46],
[0.54, 0.06, 0.74, 0.86], [0.95, 0.62, 0.28, 0.01],
[0.35, 0.71, 0.01, 0.32], [0.62, 0.24, 0.77, 0.17],
[0.73, 0.65, 0.23, 0.02], [0.27, 0.38, 0.76, 0.63],
[0.9, 0.63, 0.83, 0.6], [0.7, 0.04, 0.7, 0.82],
[0.95, 0.83, 0.64, 0.5], [0.41, 0.11, 0.61, 0.78],
[0.22, 0.44, 0.67, 0.99], [0.51, 0.05, 0.95, 0.66],
[0.99, 0.68, 0.8, 0.42], [0.72, 0.55, 0.1, 0.17],
[0.44, 0.1, 0.61, 0.98], [0.31, 0.16, 0.95, 0.9],
[0.61, 0.42, 0.24, 0.33], [0.89, 0.72, 0.78, 0.38],
[0.5, 0.09, 0.84, 0.78], [0.62, 0.01, 0.88, 0.1],
[0.44, 0.28, 0.88, 0.99], [0.57, 0.23, 0.6, 0.85],
[0.9, 0.05, 0.34, 0.41], [0.9, 0.41, 0.27, 0.36],
[0.67, 0.32, 0.66, 0.2], [0.72, 0.14, 0.63, 0.37],
[0.39, 0.08, 0.77, 0.96], [0.9, 0.7, 0.74, 0.63],
[0.63, 0.05, 0.52, 0.63], [0.62, 0.27, 0.67, 0.77],
[0.35, 0.04, 0.85, 0.86], [0.36, 0.34, 0.75, 0.37]]
centroid0 = [
0.3987179487179487, 0.7097435897435899, 0.6864102564102561,
0.4164102564102565
]
centroid1 = [
0.7788888888888889, 0.8555555555555555, 0.19666666666666668,
0.788888888888889
]
centroid2 = [
0.6038461538461538, 0.29865384615384616, 0.6217307692307692,
0.5455769230769231
]
cornelltest.assert_float_lists_equal(centroid0, cluster0.getCentroid())
cornelltest.assert_float_lists_equal(centroid1, cluster1.getCentroid())
cornelltest.assert_float_lists_equal(centroid2, cluster2.getCentroid())
cornelltest.assert_float_lists_equal(contents0, cluster0.getContents())
cornelltest.assert_float_lists_equal(contents1, cluster1.getContents())
cornelltest.assert_float_lists_equal(contents2, cluster2.getContents())
print ' Method ClusterGroup.step looks okay'
print ' Part C of class ClusterGroup appears correct'
print ''
def test_kmeans_b():
"""Test Part B of the ClusterGroup class."""
# This function tests the methods _nearest_cluster and _partition,
# both of which are private methods. Normally it's not good form to
# directly call these methods from outside the class, but we make an
# exception for testing code, which often has to be more tightly
# integrated with the implementation of a class than other code that
# just uses the class.
print ' Testing Part B of class ClusterGroup'
# Reinitialize data set
items = [[0., 0.], [10., 1.], [10., 10.], [0., 9.]]
dset = a6.Dataset(2, items)
km1 = a6.ClusterGroup(dset, 2, [0, 2])
km2 = a6.ClusterGroup(dset, 3, [0, 2, 3])
nearest = km1._nearest_cluster([1., 1.])
cornelltest.assert_true(nearest is km1.getClusters()[0])
nearest = km1._nearest_cluster([1., 10.])
cornelltest.assert_true(nearest is km1.getClusters()[1])
nearest = km2._nearest_cluster([1., 1.])
cornelltest.assert_true(nearest is km2.getClusters()[0])
nearest = km2._nearest_cluster([1., 10.])
cornelltest.assert_true(nearest is km2.getClusters()[2])
print ' Method ClusterGroup._nearest_cluster() looks okay'
# Testing partition()
# For this example points 0 and 3 are closer, as are 1 and 2
km1._partition()
cornelltest.assert_equals(set([0, 3]),
set(km1.getClusters()[0].getIndices()))
cornelltest.assert_equals(set([1, 2]),
set(km1.getClusters()[1].getIndices()))
# partition and repeat -- should not change clusters.
km1._partition()
cornelltest.assert_equals(set([0, 3]),
set(km1.getClusters()[0].getIndices()))
cornelltest.assert_equals(set([1, 2]),
set(km1.getClusters()[1].getIndices()))
# Reset the cluster centroids; now it changes
cluster = km1.getClusters()
cluster[0]._centroid = [5.0, 10.0]
cluster[1]._centroid = [0.0, 2.0]
km1._partition()
cornelltest.assert_equals(set([2, 3]),
set(km1.getClusters()[0].getIndices()))
cornelltest.assert_equals(set([0, 1]),
set(km1.getClusters()[1].getIndices()))
# Try it on a file
index1 = [
2, 3, 5, 9, 11, 15, 16, 18, 19, 20, 22, 23, 29, 30, 32, 33, 37, 40, 41,
42, 44, 45, 50, 60, 61, 62, 64, 69, 71, 73, 75, 76, 78, 80, 85, 88, 90,
94, 97
]
index2 = [0, 34, 8, 43, 66, 46, 77, 84, 54]
index3 = [
1, 4, 6, 7, 10, 12, 13, 14, 17, 21, 24, 25, 26, 27, 28, 31, 35, 36, 38,
39, 47, 48, 49, 51, 52, 53, 55, 56, 57, 58, 59, 63, 65, 67, 68, 70, 72,
74, 79, 81, 82, 83, 86, 87, 89, 91, 92, 93, 95, 96, 98, 99
]
km3 = candy_to_kmeans('datasets/smallcandy.csv', 3, [23, 54, 36])
km3._partition()
cornelltest.assert_equals(set(index1),
set(km3.getClusters()[0].getIndices()))
cornelltest.assert_equals(set(index2),
set(km3.getClusters()[1].getIndices()))
cornelltest.assert_equals(set(index3),
set(km3.getClusters()[2].getIndices()))
print ' Method ClusterGroup._partition() looks okay'
print ' Part B of class ClusterGroup appears correct'
print ''
