def runClustering(weights, k, featureFunctionMapping, businesses, truthPairs,
truthIds, cache):
cacheId = paramId(weights, k)
if (cacheId in cache):
return cache[cacheId]
featureDistMap = featureDistanceMap.FeatureDistanceMap(
featureFunctionMapping, weights)
kMeans = clustering.KMeans(k, featureDistMap)
id = "\t".join([str(weight) for weight in weights])
randIndex = -1
try:
clusters = kMeans.cluster(businesses)
randIndex = metrics.randIndex(clusters, businesses, truthPairs,
truthIds)
print("%s\t%f" % (id, randIndex), file=sys.stderr)
'''
for i in range(len(clusters)):
print("Cluster: %02d, Size: %02d" % (i, len(clusters[i])))
print(" %s" % (", ".join([str(x) for x in sorted([businesses[index].otherInfo['yelpId'] for index in clusters[i]])])))
'''
except Exception as ex:
print(ex)
print("%s\tERROR" % (id), file=sys.stderr)
cache[cacheId] = randIndex
return randIndex
def run():
businesses = features.getBusinesses(data.DATA_SOURCE_HUMAN_EVAL)
featureDistMap = featureDistanceMap.FeatureDistanceMap(learnWeights.getFeatureMapping(), WEIGHTS)
kMeans = clustering.KMeans(K, featureDistMap)
# kMeans = clustering.KMeans(K, featureDistanceMap.FeatureDistanceMap())
clusters = kMeans.cluster(businesses)
for i in range(len(clusters)):
print("Cluster: %02d, Size: %02d" % (i, len(clusters[i])))
print(" %s" % (", ".join([str(x) for x in sorted([businesses[index].otherInfo['name'] for index in clusters[i]])])))
# Metrics
truthPairs, truthIds = metrics.getHumanTruthPairs()
print("Rand Index: %f" % (metrics.randIndex(clusters, businesses, truthPairs, truthIds)))
def run(businessType):
businesses = features.getBusinesses(businessType)
# Arbitrary K
kMeans = clustering.KMeans(K, featureDistanceMap.FeatureDistanceMap())
clusters = kMeans.cluster(businesses)
for i in range(len(clusters)):
print("Cluster: %02d, Size: %02d" % (i, len(clusters[i])))
print(" %s" % (", ".join([
str(x) for x in sorted(
[businesses[index].otherInfo['name'] for index in clusters[i]])
])))
# Metrics
goldLabel = metrics.readGoldLabel("../data/groundtruth")
b_cluster = metrics.getClusterBusinessID(businesses, clusters)
randIndex = metrics.oldRandIndex(b_cluster, goldLabel)
print("Old Rand Index: " + str(randIndex))
print("New Rand Index: %f" % (metrics.randIndex(clusters, businesses)))
def run(weights, k, scalarNorm, setDistance):
businesses = features.getBusinesses(data.DATA_SOURCE_GROUNDTRUTH_ALL)
featureDistMap = featureDistanceMap.FeatureDistanceMap(
buildFeatureMapping(scalarNorm, setDistance), weights)
kMeans = clustering.KMeans(k, featureDistMap)
clusters = kMeans.cluster(businesses)
for i in range(len(clusters)):
print("Cluster: %02d, Size: %02d" % (i, len(clusters[i])))
print(" %s" % (", ".join([
str(x) for x in sorted([
businesses[index].otherInfo['yelpId'] for index in clusters[i]
])
])))
goldLabel = metrics.readGoldLabel("../data/groundtruth")
b_cluster = metrics.getClusterBusinessID(businesses, clusters)
randIndex = metrics.oldRandIndex(b_cluster, goldLabel)
return randIndex
def test_kmeansBase(self):
data = [
business.Business(10, [0, 0, 0]),
business.Business(20, [1, 1, 1]),
business.Business(30, [2, 2, 2]),
business.Business(411, [10, 10, 10]),
business.Business(511, [11, 11, 11]),
business.Business(611, [12, 12, 12]),
business.Business(7123, [110, 110, 110]),
business.Business(8123, [111, 111, 111]),
business.Business(9123, [112, 112, 112])
]
expected = [
[0, 1, 2],
[3, 4, 5],
[6, 7, 8]
]
manhattan = lambda a, b: distance.manhattan([a], [b])
kMeans = clustering.KMeans(3, featureDistanceMap.FeatureDistanceMap([manhattan, manhattan, manhattan]))
self.assertEqual(sorted(kMeans.cluster(data)), sorted(expected))
def get(self, i, j):
if (i == j):
return 0
small = min(i, j)
big = max(i, j)
return self._distances[self._calcIndex(small, big)]
if __name__ == '__main__':
data = [
business.Business(1, [0, 0, 0]),
business.Business(2, [1, 1, 1]),
business.Business(3, [2, 2, 2]),
business.Business(4, [10, 10, 10]),
business.Business(5, [11, 11, 11]),
business.Business(6, [12, 12, 12]),
business.Business(7, [110, 110, 110]),
business.Business(8, [111, 111, 111]),
business.Business(9, [112, 112, 112])
]
manhattan = lambda a, b: distance.manhattan([a], [b])
kMeans = KMeans(3, featureDistanceMap.FeatureDistanceMap([manhattan, manhattan, manhattan]))
clusters = kMeans.cluster(data)
for i in range(len(clusters)):
print("Cluster: %02d, Size: %02d - %s" % (i, len(clusters[i]), [str(x) for x in sorted(clusters[i])]))