Python feature.IDF類代碼示例

def fit_kmeans(spark, products_df):
    step = 0

    step += 1
    tokenizer = Tokenizer(inputCol="title", outputCol=str(step) + "_tokenizer")

    step += 1
    stopwords = StopWordsRemover(inputCol=tokenizer.getOutputCol(), outputCol=str(step) + "_stopwords")

    step += 1
    tf = HashingTF(inputCol=stopwords.getOutputCol(), outputCol=str(step) + "_tf", numFeatures=16)

    step += 1
    idf = IDF(inputCol=tf.getOutputCol(), outputCol=str(step) + "_idf")

    step += 1
    normalizer = Normalizer(inputCol=idf.getOutputCol(), outputCol=str(step) + "_normalizer")

    step += 1
    kmeans = KMeans(featuresCol=normalizer.getOutputCol(), predictionCol=str(step) + "_kmeans", k=2, seed=20)

    kmeans_pipeline = Pipeline(stages=[tokenizer, stopwords, tf, idf, normalizer, kmeans])

    model = kmeans_pipeline.fit(products_df)
    words_prediction = model.transform(products_df)
    model.save("./kmeans")  # the whole machine learning instance is saved in a folder
    return model, words_prediction

def tf_idf_feature(wordsData):
    hashingTF = HashingTF(inputCol="filtered", outputCol="rawFeatures", numFeatures=20)
    featurizedData = hashingTF.transform(wordsData)
    idf = IDF(inputCol="rawFeatures", outputCol="features")
    idfModel = idf.fit(featurizedData)
    rescaledData = idfModel.transform(featurizedData)
    for features_label in rescaledData.select("features", "id").take(3):
        print(features_label)

def textPredict(request):
    """6.文本聚類，熱度預測"""
    label = request.POST['label']
    title = request.POST['title']

    conf = SparkConf().setAppName('textPredict').setMaster('spark://HP-Pavilion:7077')
    sc = SparkContext(conf=conf)
    sqlContext = SQLContext(sc)
    """處理數據集，生成特征向量"""
    dfTitles = sqlContext.read.parquet('data/roll_news_sina_com_cn.parquet')
    print(dfTitles.dtypes)
    tokenizer = Tokenizer(inputCol="title", outputCol="words")
    wordsData = tokenizer.transform(dfTitles)
    hashingTF = HashingTF(inputCol="words", outputCol="rawFeatures", numFeatures=20)
    featurizedData = hashingTF.transform(wordsData)
    idf = IDF(inputCol="rawFeatures", outputCol="features")
    idfModel = idf.fit(featurizedData)
    rescaledData = idfModel.transform(featurizedData)
    rescaledData.show()
    for features_label in rescaledData.select("features", "rawFeatures").take(3):
        print(features_label)
    """決策樹模型培訓"""
    labelIndexer = StringIndexer(inputCol="label", outputCol="indexedLabel").fit(rescaledData)
    featureIndexer =\
        VectorIndexer(inputCol="features", outputCol="indexedFeatures", maxCategories=4).fit(rescaledData)
    (trainingData, testData) = rescaledData.randomSplit([0.7, 0.3])
    dt = DecisionTreeClassifier(labelCol="indexedLabel", featuresCol="indexedFeatures")
    pipeline = Pipeline(stages=[labelIndexer, featureIndexer, dt])
    model = pipeline.fit(trainingData)
    """模型測試"""
    predictions = model.transform(testData)
    predictions.show()
    predictions.select("prediction", "indexedLabel", "features").show(5)
    """用戶數據測試，單個新聞測試"""
    sentenceData = sqlContext.createDataFrame([
        (label,title),
    ],['label',"title"])
    tokenizer = Tokenizer(inputCol="title", outputCol="words")
    wordsData = tokenizer.transform(sentenceData)
    hashingTF = HashingTF(inputCol="words", outputCol="rawFeatures", numFeatures=20)
    featurizedData = hashingTF.transform(wordsData)
    rescaledData = idfModel.transform(featurizedData)
    myprediction = model.transform(rescaledData)
    print("==================================================")
    myprediction.show()
    resultList = convertDfToList(myprediction)

    """模型評估"""
    evaluator = MulticlassClassificationEvaluator(
        labelCol="indexedLabel", predictionCol="prediction", metricName="precision")
    accuracy = evaluator.evaluate(predictions)
    print("Test Error = %g " % (1.0 - accuracy))

    treeModel = model.stages[2]
    print(treeModel)

    sc.stop()
    return render(request,{'resultList':resultList})

def extract_idf_features(p_df, input_col, output_col):
    """
    Extracts IDF features.
    :param p_df: A DataFrame.
    :param in_column: Name of the input column.
    :param out_column: Name of the output column.
    :return: A DataFrame.    
    """    
    idf = IDF(inputCol=input_col, outputCol=output_col)
    idfModel = idf.fit(p_df)
    return idfModel.transform(p_df)

def run_tf_idf_spark_ml(df, numFeatures=1 << 20):
    tokenizer = Tokenizer(inputCol="body", outputCol="words")
    wordsData = tokenizer.transform(df)

    hashingTF = HashingTF(inputCol="words", outputCol="rawFeatures", numFeatures=numFeatures)
    featurizedData = hashingTF.transform(wordsData)

    idf = IDF(inputCol="rawFeatures", outputCol="features")
    idfModel = idf.fit(featurizedData)

    return idfModel.transform(featurizedData)

def tfidf(dataframe, in_col1, out_col1, in_col2, out_col2, n):

    global idfModel
    
    hashingTF = HashingTF(inputCol=in_col1, outputCol=out_col1, numFeatures=n)
    featurizedData = hashingTF.transform(dataframe)
    idf = IDF(inputCol=in_col2, outputCol=out_col2)
    idfModel = idf.fit(featurizedData)
    dataframe = idfModel.transform(featurizedData)
    
    return dataframe

def tf_feature_vectorizer(df,no_of_features,ip_col):
    #from pyspark.sql.functions import udf
    #from pyspark.sql.types import *
    output_raw_col = ip_col+"raw_features"
    output_col = ip_col+"features"
    hashingTF = HashingTF(inputCol=ip_col, outputCol=output_raw_col, numFeatures=no_of_features)
    featurizedData = hashingTF.transform(df)
    idf = IDF(inputCol=output_raw_col, outputCol=output_col)
    idfModel = idf.fit(featurizedData)
    rescaled_data = idfModel.transform(featurizedData)
    rescaled_data.show(5)
    print(rescaled_data.count())
    return rescaled_data

def create_features(raw_data):
    #Create DataFrame
    data_df = sqlContext.createDataFrame(raw_data.map(lambda r : Row(appid=r[0], price=r[1], sentence=r[2])))
    #Transform sentence into words
    tokenizer = Tokenizer(inputCol='sentence', outputCol='words')
    words_df = tokenizer.transform(data_df)
    #Calculate term frequency
    hashingTF = HashingTF(inputCol='words', outputCol='rawFeatures', numFeatures=5)
    featurized_df = hashingTF.transform(words_df)
    #Calculate inverse document frequency
    idf = IDF(inputCol='rawFeatures', outputCol='features')
    idfModel = idf.fit(featurized_df)
    return idfModel.transform(featurized_df)

 def test_idf(self):
     dataset = self.spark.createDataFrame([
         (DenseVector([1.0, 2.0]),),
         (DenseVector([0.0, 1.0]),),
         (DenseVector([3.0, 0.2]),)], ["tf"])
     idf0 = IDF(inputCol="tf")
     self.assertListEqual(idf0.params, [idf0.inputCol, idf0.minDocFreq, idf0.outputCol])
     idf0m = idf0.fit(dataset, {idf0.outputCol: "idf"})
     self.assertEqual(idf0m.uid, idf0.uid,
                      "Model should inherit the UID from its parent estimator.")
     output = idf0m.transform(dataset)
     self.assertIsNotNone(output.head().idf)
     # Test that parameters transferred to Python Model
     check_params(self, idf0m)

def makeTFIDF(sc, spark, reviews):
    # count vectorizer and tfidf
    # cv = CountVectorizer(inputCol='words_clean', outputCol='tf')
    # cvModel = cv.fit(reviews)
    # reviews = cvModel.transform(reviews)

    # HashingTF for fewer dimensions:
    hashingtf = HashingTF(inputCol='words_clean', outputCol='tf', numFeatures=1000)
    reviews = hashingtf.transform(reviews)

    # create TF-IDF matrix
    idf = IDF().setInputCol('tf').setOutputCol('tfidf')
    tfidfModel = idf.fit(reviews)
    reviews = tfidfModel.transform(reviews)

def tf_idf(df, column):
    """
    Compute TF-IDF of a corpus.
    Transformation: array<string> --> vector
    """ 
    df = preprocess(df, column) # text to list of terms
    (df, voc) = count(df, column)
    
    # creates a TF-IDF model and uses it to compute the feature vector.
    idf = IDF(inputCol=column, outputCol='_'+column)
    model = idf.fit(df)
    df = model.transform(df)
    
    df = replace(df, column, '_'+column)
    return (df, voc)

 def append_tf_idf(self, df):
     """
     Calculate term frequency and inverse document frequency
      based on at least 1 visit hourly in this case. Compares how often the tokens appeared
      at least once per hour compared to other tokens. Not used for the main purpose of the project.
     Args:
         :param df: Dataframe parameter.
     Returns:
         :return:  Dataframe with term frequency and inverse document frequency added in the columns
                     'rawFeatures' and 'features' respectively.
     """
     #Create TF column.
     hashingTF = HashingTF(inputCol="tokens", outputCol="rawFeatures", numFeatures=100000)
     tf = hashingTF.transform(df)
     tf.persist(StorageLevel.MEMORY_AND_DISK)
     #Create IDF column.
     idf = IDF(inputCol="rawFeatures", outputCol="features")
     idfModel = idf.fit(tf)
     tfidf = idfModel.transform(tf)
     return tfidf

def get_top_words(dataset, signatures):
    # TODO: Use stemmers for the languages supported by http://www.nltk.org/api/nltk.stem.html#nltk.stem.snowball.SnowballStemmer
    # Or translate comments in other languages using the free Microsoft Translate API.
    sentenceData = dataset.filter(dataset['user_comments'].isNotNull() & (dataset['useragent_locale'].isNull() | (functions.instr(dataset['useragent_locale'], 'en') == 1)))

    if sentenceData.rdd.isEmpty():
        return dict()

    # Tokenize comments.
    tokenizer = Tokenizer(inputCol='user_comments', outputCol='words')
    wordsData = tokenizer.transform(sentenceData)

    # Remove duplicate words from comments.
    wordsData = wordsData.rdd.map(lambda p: (p['signature'], list(set(p['words'])))).reduceByKey(lambda x, y: x + y).toDF(['signature', 'words'])

    if wordsData.rdd.isEmpty():
        print("[WARNING]: wordsData is empty, sentenceData wasn't.")
        return dict()

    # Clean comment words by removing puntuaction and stemming.
    def clean_word(w):
        return re.sub('\,|\.|\;|\:|\;|\?|\!|\[|\]|\}|\{|\/|\\\\', '', stem(w.lower()))

    wordsData = wordsData.rdd.map(lambda p: (p['signature'], [clean_word(w) for w in p['words']])).toDF(['signature', 'words'])

    # XXX: Useless with TF-IDF?
    remover = StopWordsRemover(inputCol='words', outputCol='filtered')
    cleanWordsData = remover.transform(wordsData)

    cv = CountVectorizer(inputCol='filtered', outputCol='features')
    model = cv.fit(cleanWordsData)
    featurizedData = model.transform(cleanWordsData)

    idf = IDF(inputCol='features', outputCol='tfidf_features')
    idfModel = idf.fit(featurizedData)
    rescaledData = idfModel.transform(featurizedData)

    bests_per_doc = rescaledData.filter(rescaledData.signature.isin(signatures)).rdd.map(lambda p: (p['signature'], sorted(zip(p['tfidf_features'].indices, p['tfidf_features'].values), key=lambda i: i[1], reverse=True)[:10])).collect()

    return dict([(signature, [model.vocabulary[best] for best, val in bests]) for signature, bests in bests_per_doc])

#print(data.head(5))
    
##creating rdd file
sc = SparkContext("local", "app")
sqc = SQLContext(sc)
df = sqc.createDataFrame(data, ['type', 'text'])

#NEW VARIABLE GENERATION
dataCleaned = df.map(lambda x: (1 if x['type'] == 'spam' else 0, tokenize(x['text'])))
dataClean = dataCleaned.map(lambda x: (float(x[0]), x[1]))
dfClean = sqc.createDataFrame(dataClean, ['label', 'words'])
dfClean.show(5)

hashingTF = HashingTF(inputCol="words", outputCol="rawtf-idf", numFeatures=1000)
tf = hashingTF.transform(dfClean)
idf = IDF(inputCol="rawtf-idf", outputCol="features").fit(tf)
dfFinal = idf.transform(tf)

# Fit on whole dataset to include all labels in index.
labelIndexer = StringIndexer(inputCol="label", outputCol="indexedLabel").fit(dfFinal)
# Automatically identify categorical features, and index them.
# Set maxCategories so features with > 4 distinct values are treated as continuous.
featureIndexer = VectorIndexer(inputCol="features", outputCol="indexedFeatures", maxCategories=4).fit(dfFinal)

# Split the data into training and test sets (20% held out for testing)
(trainingData, testData) = dfFinal.randomSplit([0.8, 0.2])


# Train the model.
#rf = RandomForestClassifier(labelCol="indexedLabel", featuresCol="indexedFeatures")
nb = NaiveBayes(smoothing = 1.0, labelCol="indexedLabel", featuresCol="indexedFeatures")

    review_text = BeautifulSoup(raw_review).text
    #
    # 2. Remove non-letters        
    letters_only = re.sub("[^a-zA-Z]", " ", review_text) 
    #
    # 3. Convert to lower case, split into individual words
    words = letters_only.lower().split()                                            
    # 
    # 4. Remove stop words
    meaningful_words =  [w for w in words if not w in stops]   
    #
    # 5. Join the words back into one string separated by space, 
    # and return the result.
    return " ".join( meaningful_words)   

stops = set(stopwords.words("english")) 
lines = sc.textFile("s3://spark-project-data/labeledTrainData.tsv")
rows = lines.zipWithIndex().filter(lambda (row,index): index > 0).keys()
parts = rows.map(lambda l: l.split("\t"))

review = parts.map(lambda p: Row(id=p[0], label=float(p[1]), 
	review=review_to_words(p[2])))
schemeReview = sqlContext.createDataFrame(review)
tokenizer = Tokenizer(inputCol="review", outputCol="words")
wordsData = tokenizer.transform(schemeReview)
hashingTF = HashingTF(inputCol="words", outputCol="rawFeatures", numFeatures=300)
featurizedData = hashingTF.transform(wordsData)
idf = IDF(inputCol="rawFeatures", outputCol="features")
idfModel = idf.fit(featurizedData)
rescaledData = idfModel.transform(featurizedData)
selectData = rescaledData.select("label","features")