spark 数据预处理 特征标准化 归一化模块

#We will also standardise our data as we have done so far when performing distance-based clustering.

from pyspark.mllib.feature import StandardScaler
standardizer = StandardScaler(True, True)
t0 = time()
standardizer_model = standardizer.fit(parsed_data_values)
tt = time() - t0
standardized_data_values = standardizer_model.transform(parsed_data_values)
print "Data standardized in {} seconds".format(round(tt,3))

Data standardized in 9.54 seconds

We can now perform k-means clustering.

from pyspark.mllib.clustering import KMeans
t0 = time()
clusters = KMeans.train(standardized_data_values, 80,
                        maxIterations=10, runs=5,
                        initializationMode="random")
tt = time() - t0
print "Data clustered in {} seconds".format(round(tt,3))

Data clustered in 137.496 seconds

kmeans demo

摘自：http://spark.apache.org/docs/latest/api/python/pyspark.mllib.html#module-pyspark.mllib.feature

pyspark.mllib.feature module

Python package for feature in MLlib.

class pyspark.mllib.feature.Normalizer(p=2.0)[source]

Bases: pyspark.mllib.feature.VectorTransformer

Normalizes samples individually to unit L^p norm

For any 1 <= p < float(‘inf’), normalizes samples using sum(abs(vector) ^p) ^(1/p) as norm.

For p = float(‘inf’), max(abs(vector)) will be used as norm for normalization.

Parameters:	p – Normalization in L^p^ space, p = 2 by default.

>>> v = Vectors.dense(range(3))
>>> nor = Normalizer(1)
>>> nor.transform(v)
DenseVector([0.0, 0.3333, 0.6667])

>>> rdd = sc.parallelize([v])
>>> nor.transform(rdd).collect()
[DenseVector([0.0, 0.3333, 0.6667])]

>>> nor2 = Normalizer(float("inf"))
>>> nor2.transform(v)
DenseVector([0.0, 0.5, 1.0])

New in version 1.2.0.

transform(vector)[source]

Applies unit length normalization on a vector.

Parameters:	vector – vector or RDD of vector to be normalized.
Returns:	normalized vector. If the norm of the input is zero, it will return the input vector.

New in version 1.2.0.

class pyspark.mllib.feature.StandardScalerModel(java_model)[source]

Bases: pyspark.mllib.feature.JavaVectorTransformer

Represents a StandardScaler model that can transform vectors.

New in version 1.2.0.

mean[source]

Return the column mean values.

New in version 2.0.0.

setWithMean(withMean)[source]

Setter of the boolean which decides whether it uses mean or not

New in version 1.4.0.

setWithStd(withStd)[source]

Setter of the boolean which decides whether it uses std or not

New in version 1.4.0.

std[source]

Return the column standard deviation values.

New in version 2.0.0.

transform(vector)[source]

Applies standardization transformation on a vector.

Note

In Python, transform cannot currently be used within an RDD transformation or action. Call transform directly on the RDD instead.

Parameters:	vector – Vector or RDD of Vector to be standardized.
Returns:	Standardized vector. If the variance of a column is zero, it will return default 0.0 for the column with zero variance.

New in version 1.2.0.

withMean[source]

Returns if the model centers the data before scaling.

New in version 2.0.0.

withStd[source]

Returns if the model scales the data to unit standard deviation.

New in version 2.0.0.

class pyspark.mllib.feature.StandardScaler(withMean=False, withStd=True)[source]

Bases: object

Standardizes features by removing the mean and scaling to unit variance using column summary statistics on the samples in the training set.

Parameters:	withMean – False by default. Centers the data with mean before scaling. It will build a dense output, so take care when applying to sparse input. withStd – True by default. Scales the data to unit standard deviation.

>>> vs = [Vectors.dense([-2.0, 2.3, 0]), Vectors.dense([3.8, 0.0, 1.9])]
>>> dataset = sc.parallelize(vs)
>>> standardizer = StandardScaler(True, True)
>>> model = standardizer.fit(dataset)
>>> result = model.transform(dataset)
>>> for r in result.collect(): r
DenseVector([-0.7071, 0.7071, -0.7071])
DenseVector([0.7071, -0.7071, 0.7071])
>>> int(model.std[0])
4
>>> int(model.mean[0]*10)
9
>>> model.withStd
True
>>> model.withMean
True

New in version 1.2.0.

fit(dataset)[source]

Computes the mean and variance and stores as a model to be used for later scaling.

Parameters:	dataset – The data used to compute the mean and variance to build the transformation model.
Returns:	a StandardScalarModel

New in version 1.2.0.

class pyspark.mllib.feature.HashingTF(numFeatures=1048576)[source]

Bases: object

Maps a sequence of terms to their term frequencies using the hashing trick.

Note

The terms must be hashable (can not be dict/set/list...).

Parameters:	numFeatures – number of features (default: 2^20)

>>> htf = HashingTF(100)
>>> doc = "a a b b c d".split(" ")
>>> htf.transform(doc)
SparseVector(100, {...})

New in version 1.2.0.

indexOf(term)[source]

Returns the index of the input term.

New in version 1.2.0.

setBinary(value)[source]

If True, term frequency vector will be binary such that non-zero term counts will be set to 1 (default: False)

New in version 2.0.0.

transform(document)[source]

Transforms the input document (list of terms) to term frequency vectors, or transform the RDD of document to RDD of term frequency vectors.

New in version 1.2.0.

class pyspark.mllib.feature.IDFModel(java_model)[source]

Bases: pyspark.mllib.feature.JavaVectorTransformer

Represents an IDF model that can transform term frequency vectors.

New in version 1.2.0.

idf()[source]

Returns the current IDF vector.

New in version 1.4.0.

transform(x)[source]

Transforms term frequency (TF) vectors to TF-IDF vectors.

If minDocFreq was set for the IDF calculation, the terms which occur in fewer than minDocFreq documents will have an entry of 0.

Note

In Python, transform cannot currently be used within an RDD transformation or action. Call transform directly on the RDD instead.

Parameters:	x – an RDD of term frequency vectors or a term frequency vector
Returns:	an RDD of TF-IDF vectors or a TF-IDF vector

New in version 1.2.0.

class pyspark.mllib.feature.IDF(minDocFreq=0)[source]

Bases: object

Inverse document frequency (IDF).

The standard formulation is used: idf = log((m + 1) / (d(t) + 1)), where m is the total number of documents and d(t) is the number of documents that contain term t.

This implementation supports filtering out terms which do not appear in a minimum number of documents (controlled by the variable minDocFreq). For terms that are not in at least minDocFreq documents, the IDF is found as 0, resulting in TF-IDFs of 0.

Parameters:	minDocFreq – minimum of documents in which a term should appear for filtering

>>> n = 4
>>> freqs = [Vectors.sparse(n, (1, 3), (1.0, 2.0)),
...          Vectors.dense([0.0, 1.0, 2.0, 3.0]),
...          Vectors.sparse(n, [1], [1.0])]
>>> data = sc.parallelize(freqs)
>>> idf = IDF()
>>> model = idf.fit(data)
>>> tfidf = model.transform(data)
>>> for r in tfidf.collect(): r
SparseVector(4, {1: 0.0, 3: 0.5754})
DenseVector([0.0, 0.0, 1.3863, 0.863])
SparseVector(4, {1: 0.0})
>>> model.transform(Vectors.dense([0.0, 1.0, 2.0, 3.0]))
DenseVector([0.0, 0.0, 1.3863, 0.863])
>>> model.transform([0.0, 1.0, 2.0, 3.0])
DenseVector([0.0, 0.0, 1.3863, 0.863])
>>> model.transform(Vectors.sparse(n, (1, 3), (1.0, 2.0)))
SparseVector(4, {1: 0.0, 3: 0.5754})

New in version 1.2.0.

fit(dataset)[source]

Computes the inverse document frequency.

Parameters:	dataset – an RDD of term frequency vectors

New in version 1.2.0.

class pyspark.mllib.feature.Word2Vec[source]

Bases: object

Word2Vec creates vector representation of words in a text corpus. The algorithm first constructs a vocabulary from the corpus and then learns vector representation of words in the vocabulary. The vector representation can be used as features in natural language processing and machine learning algorithms.

We used skip-gram model in our implementation and hierarchical softmax method to train the model. The variable names in the implementation matches the original C implementation.

For original C implementation, see https://code.google.com/p/word2vec/ For research papers, see Efficient Estimation of Word Representations in Vector Space and Distributed Representations of Words and Phrases and their Compositionality.

>>> sentence = "a b " * 100 + "a c " * 10
>>> localDoc = [sentence, sentence]
>>> doc = sc.parallelize(localDoc).map(lambda line: line.split(" "))
>>> model = Word2Vec().setVectorSize(10).setSeed(42).fit(doc)

Querying for synonyms of a word will not return that word:

>>> syms = model.findSynonyms("a", 2)
>>> [s[0] for s in syms]
[u'b', u'c']

But querying for synonyms of a vector may return the word whose representation is that vector:

>>> vec = model.transform("a")
>>> syms = model.findSynonyms(vec, 2)
>>> [s[0] for s in syms]
[u'a', u'b']

>>> import os, tempfile
>>> path = tempfile.mkdtemp()
>>> model.save(sc, path)
>>> sameModel = Word2VecModel.load(sc, path)
>>> model.transform("a") == sameModel.transform("a")
True
>>> syms = sameModel.findSynonyms("a", 2)
>>> [s[0] for s in syms]
[u'b', u'c']
>>> from shutil import rmtree
>>> try:
...     rmtree(path)
... except OSError:
...     pass

New in version 1.2.0.

fit(data)[source]

Computes the vector representation of each word in vocabulary.

Parameters:	data – training data. RDD of list of string
Returns:	Word2VecModel instance

New in version 1.2.0.

setLearningRate(learningRate)[source]

Sets initial learning rate (default: 0.025).

New in version 1.2.0.

setMinCount(minCount)[source]

Sets minCount, the minimum number of times a token must appear to be included in the word2vec model’s vocabulary (default: 5).

New in version 1.4.0.

setNumIterations(numIterations)[source]

Sets number of iterations (default: 1), which should be smaller than or equal to number of partitions.

New in version 1.2.0.

setNumPartitions(numPartitions)[source]

Sets number of partitions (default: 1). Use a small number for accuracy.

New in version 1.2.0.

setSeed(seed)[source]

Sets random seed.

New in version 1.2.0.

setVectorSize(vectorSize)[source]

Sets vector size (default: 100).

New in version 1.2.0.

setWindowSize(windowSize)[source]

Sets window size (default: 5).

New in version 2.0.0.

class pyspark.mllib.feature.Word2VecModel(java_model)[source]

Bases: pyspark.mllib.feature.JavaVectorTransformer, pyspark.mllib.util.JavaSaveable, pyspark.mllib.util.JavaLoader

class for Word2Vec model

New in version 1.2.0.

findSynonyms(word, num)[source]

Find synonyms of a word

Parameters:	word – a word or a vector representation of word num – number of synonyms to find
Returns:	array of (word, cosineSimilarity)

Note

Local use only

New in version 1.2.0.

getVectors()[source]

Returns a map of words to their vector representations.

New in version 1.4.0.

classmethod load(sc, path)[source]

Load a model from the given path.

New in version 1.5.0.

transform(word)[source]

Transforms a word to its vector representation

Note

Local use only

Parameters:	word – a word
Returns:	vector representation of word(s)

New in version 1.2.0.

class pyspark.mllib.feature.ChiSqSelector(numTopFeatures=50, selectorType='numTopFeatures', percentile=0.1, fpr=0.05, fdr=0.05, fwe=0.05)[source]

Bases: object

Creates a ChiSquared feature selector. The selector supports different selection methods: numTopFeatures, percentile, fpr, fdr, fwe.

• numTopFeatures chooses a fixed number of top features according to a chi-squared test.
• percentile is similar but chooses a fraction of all features instead of a fixed number.
• fpr chooses all features whose p-values are below a threshold, thus controlling the false positive rate of selection.
• fdr uses the Benjamini-Hochberg procedure to choose all features whose false discovery rate is below a threshold.
• fwe chooses all features whose p-values are below a threshold. The threshold is scaled by 1/numFeatures, thus controlling the family-wise error rate of selection.

By default, the selection method is numTopFeatures, with the default number of top features set to 50.

>>> data = sc.parallelize([
...     LabeledPoint(0.0, SparseVector(3, {0: 8.0, 1: 7.0})),
...     LabeledPoint(1.0, SparseVector(3, {1: 9.0, 2: 6.0})),
...     LabeledPoint(1.0, [0.0, 9.0, 8.0]),
...     LabeledPoint(2.0, [7.0, 9.0, 5.0]),
...     LabeledPoint(2.0, [8.0, 7.0, 3.0])
... ])
>>> model = ChiSqSelector(numTopFeatures=1).fit(data)
>>> model.transform(SparseVector(3, {1: 9.0, 2: 6.0}))
SparseVector(1, {})
>>> model.transform(DenseVector([7.0, 9.0, 5.0]))
DenseVector([7.0])
>>> model = ChiSqSelector(selectorType="fpr", fpr=0.2).fit(data)
>>> model.transform(SparseVector(3, {1: 9.0, 2: 6.0}))
SparseVector(1, {})
>>> model.transform(DenseVector([7.0, 9.0, 5.0]))
DenseVector([7.0])
>>> model = ChiSqSelector(selectorType="percentile", percentile=0.34).fit(data)
>>> model.transform(DenseVector([7.0, 9.0, 5.0]))
DenseVector([7.0])

New in version 1.4.0.

fit(data)[source]

Returns a ChiSquared feature selector.

Parameters:	data – an RDD[LabeledPoint] containing the labeled dataset with categorical features. Real-valued features will be treated as categorical for each distinct value. Apply feature discretizer before using this function.

New in version 1.4.0.

setFdr(fdr)[source]

set FDR [0.0, 1.0] for feature selection by FDR. Only applicable when selectorType = “fdr”.

New in version 2.2.0.

setFpr(fpr)[source]

set FPR [0.0, 1.0] for feature selection by FPR. Only applicable when selectorType = “fpr”.

New in version 2.1.0.

setFwe(fwe)[source]

set FWE [0.0, 1.0] for feature selection by FWE. Only applicable when selectorType = “fwe”.

New in version 2.2.0.

setNumTopFeatures(numTopFeatures)[source]

set numTopFeature for feature selection by number of top features. Only applicable when selectorType = “numTopFeatures”.

New in version 2.1.0.

setPercentile(percentile)[source]

set percentile [0.0, 1.0] for feature selection by percentile. Only applicable when selectorType = “percentile”.

New in version 2.1.0.

setSelectorType(selectorType)[source]

set the selector type of the ChisqSelector. Supported options: “numTopFeatures” (default), “percentile”, “fpr”, “fdr”, “fwe”.

New in version 2.1.0.

class pyspark.mllib.feature.ChiSqSelectorModel(java_model)[source]

Bases: pyspark.mllib.feature.JavaVectorTransformer

Represents a Chi Squared selector model.

New in version 1.4.0.

transform(vector)[source]

Applies transformation on a vector.

Parameters:	vector – Vector or RDD of Vector to be transformed.
Returns:	transformed vector.

New in version 1.4.0.

class pyspark.mllib.feature.ElementwiseProduct(scalingVector)[source]

Bases: pyspark.mllib.feature.VectorTransformer

Scales each column of the vector, with the supplied weight vector. i.e the elementwise product.

>>> weight = Vectors.dense([1.0, 2.0, 3.0])
>>> eprod = ElementwiseProduct(weight)
>>> a = Vectors.dense([2.0, 1.0, 3.0])
>>> eprod.transform(a)
DenseVector([2.0, 2.0, 9.0])
>>> b = Vectors.dense([9.0, 3.0, 4.0])
>>> rdd = sc.parallelize([a, b])
>>> eprod.transform(rdd).collect()
[DenseVector([2.0, 2.0, 9.0]), DenseVector([9.0, 6.0, 12.0])]

New in version 1.5.0.

transform(vector)[source]

Computes the Hadamard product of the vector.

New in version 1.5.0.