Python feature.Tokenizer类代码示例

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 token(dataframe, in_col, out_col):
    
    tokenizer = Tokenizer(inputCol=in_col, outputCol=out_col)
    dataframe = tokenizer.transform(dataframe)
    
    dataframe.printSchema()
    
    return dataframe

def textPredict(request):
    """6.文本聚类，热度预测"""
    label = request.POST['label']
    title = request.POST['title']

    conf = SparkConf().setAppName('textPredict').setMaster('spark://HP-Pavilion:7077')
    sc = SparkContext(conf=conf)
    sqlContext = SQLContext(sc)
    """处理数据集，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)
    """决策树模型培训"""
    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)
    """模型测试"""
    predictions = model.transform(testData)
    predictions.show()
    predictions.select("prediction", "indexedLabel", "features").show(5)
    """用户数据测试，单个新闻测试"""
    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)

    """模型评估"""
    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 main():
    spark = SparkSession.builder.appName("DBPediaSpark").getOrCreate()

    args = getResolvedOptions(sys.argv, ['S3_INPUT_BUCKET',
                                         'S3_INPUT_KEY_PREFIX',
                                         'S3_OUTPUT_BUCKET',
                                         'S3_OUTPUT_KEY_PREFIX',
                                         'S3_MODEL_BUCKET',
                                         'S3_MODEL_KEY_PREFIX'])

    # This is needed to save RDDs which is the only way to write nested Dataframes into CSV format
    spark.sparkContext._jsc.hadoopConfiguration().set("mapred.output.committer.class",
                                                      "org.apache.hadoop.mapred.FileOutputCommitter")
    
    # Defining the schema corresponding to the input data. The input data does not contain the headers
    schema = StructType([StructField("label", IntegerType(), True), 
                         StructField("title", StringType(), True), 
                         StructField("abstract", StringType(), True)])
    
    # Download the data from S3 into two separate Dataframes
    traindf = spark.read.csv(('s3://' + os.path.join(args['S3_INPUT_BUCKET'], args['S3_INPUT_KEY_PREFIX'],
                                                   'train.csv')), header=False, schema=schema, encoding='UTF-8')
    validationdf = spark.read.csv(('s3://' + os.path.join(args['S3_INPUT_BUCKET'], args['S3_INPUT_KEY_PREFIX'],
                                                          'test.csv')), header=False, schema=schema, encoding='UTF-8')

    # Tokenize the abstract column which contains the input text
    tokenizer = Tokenizer(inputCol="abstract", outputCol="tokenized_abstract")

    # Save transformed training data to CSV in S3 by converting to RDD.
    transformed_traindf = tokenizer.transform(traindf)
    transformed_train_rdd = transformed_traindf.rdd.map(lambda x: (x.label, x.tokenized_abstract))
    lines = transformed_train_rdd.map(csv_line)
    lines.coalesce(1).saveAsTextFile('s3://' + os.path.join(args['S3_OUTPUT_BUCKET'], args['S3_OUTPUT_KEY_PREFIX'], 'train'))

    # Similar data processing for validation dataset.
    transformed_validation = tokenizer.transform(validationdf)
    transformed_validation_rdd = transformed_validation.rdd.map(lambda x: (x.label, x.tokenized_abstract))
    lines = transformed_validation_rdd.map(csv_line)
    lines.coalesce(1).saveAsTextFile('s3://' + os.path.join(args['S3_OUTPUT_BUCKET'], args['S3_OUTPUT_KEY_PREFIX'], 'validation'))

    # Serialize the tokenizer via MLeap and upload to S3
    SimpleSparkSerializer().serializeToBundle(tokenizer, "jar:file:/tmp/model.zip", transformed_validation)

    # Unzip as SageMaker expects a .tar.gz file but MLeap produces a .zip file.
    import zipfile
    with zipfile.ZipFile("/tmp/model.zip") as zf:
        zf.extractall("/tmp/model")

    # Write back the content as a .tar.gz file
    import tarfile
    with tarfile.open("/tmp/model.tar.gz", "w:gz") as tar:
        tar.add("/tmp/model/bundle.json", arcname='bundle.json')
        tar.add("/tmp/model/root", arcname='root')

    s3 = boto3.resource('s3')
    file_name = os.path.join(args['S3_MODEL_KEY_PREFIX'], 'model.tar.gz')
    s3.Bucket(args['S3_MODEL_BUCKET']).upload_file('/tmp/model.tar.gz', file_name)

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 predictLabel(label,title,model):
    """预测新闻的标签"""
    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)
    return myprediction

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 _build_stages(self):
     self.bs_parser = BeautifulSoupParser(inputCol="review", outputCol="parsed")
     self.tokenizer = Tokenizer(inputCol=self.bs_parser.getOutputCol(), outputCol="words")
     self.hashing_tf = HashingTF(inputCol=self.tokenizer.getOutputCol(), outputCol="raw_features")
     self.idf_model = IDF(inputCol=self.hashing_tf.getOutputCol(), outputCol="features")
     self.lr = LogisticRegression(maxIter=10, regParam=0.01)
     return [self.bs_parser, self.tokenizer, self.hashing_tf, self.idf_model, self.lr]

def preprocessing_titles(path,name):
    query = preprocessData(path)
    tokenizer = Tokenizer(inputCol="title", outputCol="tokenized_title")
    wordsData = tokenizer.transform(query)
    #after Stopword removal
    remover = StopWordsRemover(inputCol="tokenized_title", outputCol="filtered")
    wordsData= remover.transform(wordsData)
    
    df = wordsData.map(lambda x:x['id']).zipWithUniqueId().toDF(["id","index"])
    df.registerTempTable("indices")
    wordsData.registerTempTable("words")
    
    qr = sqlContext.sql("SELECT index,words.id,filtered FROM indices JOIN words ON words.id = indices.id")
    if name!='':
        exportOnS3(qr,"s3a://redit-preprocessed/",name)
    qr = qr.map(lambda Row:(Row['index'],Row['id'],Row['filtered']))

    def getPipeline(self, df):
        # notify pipeline 
        self.success('Initializing ML Pipeline ...')

        # initialize our tokenizer, we're going to tokenize features
        tokenizer = Tokenizer(inputCol='tag_features', outputCol='words')
        # convert the tokenize data to vectorize data
        hashingTF = HashingTF(inputCol=tokenizer.getOutputCol(), outputCol='features')
        # initialize logistic regression algorithm
        lr        = LogisticRegression(maxIter=10, regParam=0.01)
        # create / initialize the ml pipeline
        pipeline  = Pipeline(stages=[tokenizer, hashingTF, lr])

        # fit the pipeline on our training dataframe
        model = pipeline.fit(df)

        return model

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])

def main():
    '''
    takes one input argument :: Location of the directory for training and test data files.
    :return: Print output on console for the area under the ROC curve.
    '''

    conf = SparkConf().setAppName("MLPipeline")
    sc = SparkContext(conf=conf)

    # Read training data as a DataFrame
    sqlCt = SQLContext(sc)
    trainDF = sqlCt.read.parquet("20news_train.parquet")

    # Configure an ML pipeline, which consists of three stages: tokenizer, hashingTF, and lr.
    tokenizer = Tokenizer(inputCol="text", outputCol="words")
    hashingTF = HashingTF(inputCol=tokenizer.getOutputCol(), outputCol="features", numFeatures=1000)
    lr = LogisticRegression(maxIter=20, regParam=0.1)
    pipeline = Pipeline(stages=[tokenizer, hashingTF, lr])

    # Fit the pipeline to training data.
    model = pipeline.fit(trainDF)

    numFeatures = (1000, 5000, 10000)
    regParam = (0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9)
    paramGrid = ParamGridBuilder().addGrid(hashingTF.numFeatures, numFeatures).addGrid(lr.regParam, regParam).build()


    cv = CrossValidator().setEstimator(pipeline).setEvaluator(BinaryClassificationEvaluator()).setEstimatorParamMaps(paramGrid).setNumFolds(2)

    # Evaluate the model on testing data
    testDF = sqlCt.read.parquet("20news_test.parquet")
    prediction = model.transform(testDF)
    evaluator = BinaryClassificationEvaluator()


    model_cv = cv.fit(trainDF)
    prediction_cv = model_cv.transform(testDF)
    print evaluator.evaluate(prediction)
    print evaluator.evaluate(prediction_cv)

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

    words = wordsData.select("words").rdd.map(lambda x: x.words)

    hashingTF = MllibHashingTF(numFeatures)
    tf = hashingTF.transform(words)
    tf.cache()

    idf = MllibIDF().fit(tf)
    tfidf = idf.transform(tf)

    # @TODO make this nicer
    tmp = sqlContext.createDataFrame(wordsData.rdd.zip(tfidf), ["data", "features"])
    tmp.registerTempTable("tmp")
    old_columns = ', '.join(map(lambda x: 'data.%s' % x, wordsData.columns))
    with_features = sqlContext.sql("SELECT %s, features FROM tmp" % old_columns)
    tmp = sqlContext.createDataFrame(with_features.rdd.zip(tf), ["data", "rawFeatures"])
    tmp.registerTempTable("tmp")
    old_columns = ', '.join(map(lambda x: 'data.%s' % x, with_features.columns))
    return sqlContext.sql("SELECT %s, rawFeatures FROM tmp" % old_columns)

def main():
    # Read training data as a DataFrame
    sqlCt = SQLContext(sc)
    trainDF = sqlCt.read.parquet(training_input)
    testDF = sqlCt.read.parquet(testing_input)

    tokenizer = Tokenizer(inputCol="text", outputCol="words")
    evaluator = BinaryClassificationEvaluator()

    # no parameter tuning
    hashingTF_notuning = HashingTF(inputCol=tokenizer.getOutputCol(), outputCol="features", numFeatures=1000)
    lr_notuning = LogisticRegression(maxIter=20, regParam=0.1)
    pipeline_notuning = Pipeline(stages=[tokenizer, hashingTF_notuning, lr_notuning])
    model_notuning = pipeline_notuning.fit(trainDF)

    prediction_notuning = model_notuning.transform(testDF)
    notuning_output = evaluator.evaluate(prediction_notuning)

    # for cross validation
    hashingTF = HashingTF(inputCol=tokenizer.getOutputCol(), outputCol="features")
    lr = LogisticRegression(maxIter=20)

    paramGrid = ParamGridBuilder()\
        .addGrid(hashingTF.numFeatures, [1000, 5000, 10000])\
        .addGrid(lr.regParam, [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9])\
        .build()

    pipeline = Pipeline(stages=[tokenizer, hashingTF, lr])
    cv = CrossValidator(estimator=pipeline, estimatorParamMaps=paramGrid, evaluator=evaluator, numFolds=2)
    cvModel = cv.fit(trainDF)

    # Make predictions on test documents. cvModel uses the best model found.
    best_prediction = cvModel.transform(testDF)
    best_output = evaluator.evaluate(best_prediction)

    s = str(notuning_output) + '\n' + str(best_output)
    output_data = sc.parallelize([s])
    output_data.saveAsTextFile(output)

class BaselinePipelineEngine(PipelineEngine):
    @keyword_only
    def __init__(self, cv):
        super(BaselinePipelineEngine, self).__init__(cv)
        self.hashing_tf_map = [pow(2, 20)]
        self.lr_map = [0.1, 0.01]
        self.stages = self._build_stages()
        self.pipeline = Pipeline(stages=[self.bs_parser, self.tokenizer, self.hashing_tf, self.idf_model, self.lr])
        self.param_grid = self._build_param_grid()

    def _build_stages(self):
        self.bs_parser = BeautifulSoupParser(inputCol="review", outputCol="parsed")
        self.tokenizer = Tokenizer(inputCol=self.bs_parser.getOutputCol(), outputCol="words")
        self.hashing_tf = HashingTF(inputCol=self.tokenizer.getOutputCol(), outputCol="raw_features")
        self.idf_model = IDF(inputCol=self.hashing_tf.getOutputCol(), outputCol="features")
        self.lr = LogisticRegression(maxIter=10, regParam=0.01)
        return [self.bs_parser, self.tokenizer, self.hashing_tf, self.idf_model, self.lr]

    def _build_param_grid(self):
        param_grid_builder = ParamGridBuilder()
        param_grid_builder.addGrid(self.hashing_tf.numFeatures, self.hashing_tf_map)
        param_grid_builder.addGrid(self.lr.regParam, self.lr_map)
        return param_grid_builder.build()