def testUserPoints(dataset):
algo = H2OKMeans(
splitRatio=0.8,
seed=1,
k=3,
featuresCols=["sepal_len", "sepal_wid", "petal_len", "petal_wid"],
userPoints=[[4.9, 3.0, 1.4, 0.2], [5.6, 2.5, 3.9, 1.1],
[6.5, 3.0, 5.2, 2.0]])
model = algo.fit(dataset)
assert model.transform(dataset).select(
"prediction").head()[0] == 0, "Prediction should match"
def testFullResultInPredictionDetailsCol(dataset):
algo = H2OKMeans(
splitRatio=0.8,
seed=1,
k=3,
featuresCols=["sepal_len", "sepal_wid", "petal_len", "petal_wid"])
model = algo.fit(dataset)
transformed = model.transform(dataset)
assert transformed.select("detailed_prediction.cluster").head(
)[0] == 0, "Prediction should match"
assert len(transformed.select("detailed_prediction.distances").head()
[0]) == 3, "Size of distances array should match"
def testResultInPredictionCol(dataset):
algo = H2OKMeans(
splitRatio=0.8,
seed=1,
k=3,
featuresCols=["sepal_len", "sepal_wid", "petal_len", "petal_wid"])
model = algo.fit(dataset)
transformed = model.transform(dataset)
assert transformed.select(
"prediction").head()[0] == 0, "Prediction should match"
assert transformed.select("prediction").distinct().count(
) == 3, "Number of clusters should match"
def testParams():
kmeans = H2OKMeans(predictionCol="prediction",
detailedPredictionCol="detailed_prediction",
withDetailedPredictionCol=False,
featuresCols=[],
foldCol=None,
weightCol=None,
splitRatio=1.0,
seed=-1,
nfolds=0,
allStringColumnsToCategorical=True,
columnsToCategorical=[],
convertUnknownCategoricalLevelsToNa=False,
convertInvalidNumbersToNa=False,
modelId=None,
keepCrossValidationPredictions=False,
keepCrossValidationFoldAssignment=False,
parallelizeCrossValidation=True,
distribution="AUTO",
maxIterations=10,
standardize=True,
init="Furthest",
userPoints=None,
estimateK=False,
k=2)
assert kmeans.getPredictionCol() == "prediction"
assert kmeans.getDetailedPredictionCol() == "detailed_prediction"
assert kmeans.getWithDetailedPredictionCol() == False
assert kmeans.getFeaturesCols() == []
assert kmeans.getFoldCol() == None
assert kmeans.getWeightCol() == None
assert kmeans.getSplitRatio() == 1.0
assert kmeans.getSeed() == -1
assert kmeans.getNfolds() == 0
assert kmeans.getAllStringColumnsToCategorical() == True
assert kmeans.getColumnsToCategorical() == []
assert kmeans.getConvertUnknownCategoricalLevelsToNa() == False
assert kmeans.getConvertInvalidNumbersToNa() == False
assert kmeans.getModelId() == None
assert kmeans.getKeepCrossValidationPredictions() == False
assert kmeans.getKeepCrossValidationFoldAssignment() == False
assert kmeans.getParallelizeCrossValidation() == True
assert kmeans.getDistribution() == "AUTO"
assert kmeans.getMaxIterations() == 10
assert kmeans.getStandardize() == True
assert kmeans.getInit() == "Furthest"
assert kmeans.getUserPoints() == None
assert kmeans.getEstimateK() == False
assert kmeans.getK() == 2
def testFullResultInPredictionDetailsCol(self):
algo = H2OKMeans(
splitRatio=0.8,
seed=1,
k=3,
featuresCols=["sepal_len", "sepal_wid", "petal_len", "petal_wid"],
withDetailedPredictionCol=True)
model = algo.fit(self.dataset)
transformed = model.transform(self.dataset)
self.assertEquals(
transformed.select("detailed_prediction.cluster").head()[0], 0,
"Prediction should match")
self.assertEquals(
len(transformed.select("detailed_prediction.distances").head()[0]),
3, "Size of distances array should match")
def testPipelineSerialization(dataset):
algo = H2OKMeans(
splitRatio=0.8,
seed=1,
k=3,
featuresCols=["sepal_len", "sepal_wid", "petal_len", "petal_wid"])
pipeline = Pipeline(stages=[algo])
pipeline.write().overwrite().save("file://" +
os.path.abspath("build/kmeans_pipeline"))
loadedPipeline = Pipeline.load("file://" +
os.path.abspath("build/kmeans_pipeline"))
model = loadedPipeline.fit(dataset)
model.write().overwrite().save(
"file://" + os.path.abspath("build/kmeans_pipeline_model"))
loadedModel = PipelineModel.load(
"file://" + os.path.abspath("build/kmeans_pipeline_model"))
loadedModel.transform(dataset).count()
def testPipelineSerializationKMeans(prostateDataset):
gridSearchTester(H2OKMeans(), prostateDataset)
def testKmeansParameters(prostateDataset):
features = ['AGE', 'RACE', 'DPROS', 'DCAPS', 'PSA']
algorithm = H2OKMeans(seed=1, featuresCols=features)
model = algorithm.fit(prostateDataset)
compareParameterValues(algorithm, model)