├── settings.gradle ├── gradle └── wrapper │ ├── gradle-wrapper.jar │ └── gradle-wrapper.properties ├── .gitignore ├── src └── main │ └── scala │ └── ru │ └── ispras │ └── pu4spark │ ├── ProbabilisticClassifierConfig.scala │ ├── PositiveUnlabeledLearner.scala │ ├── TraditionalPULearner.scala │ ├── TwoStepPULearner.scala │ └── GradualReductionPULearner.scala ├── gradlew.bat ├── README.md ├── gradlew └── LICENSE /settings.gradle: -------------------------------------------------------------------------------- 1 | rootProject.name = 'pu4spark' 2 | 3 | -------------------------------------------------------------------------------- /gradle/wrapper/gradle-wrapper.jar: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ispras/pu4spark/HEAD/gradle/wrapper/gradle-wrapper.jar -------------------------------------------------------------------------------- /gradle/wrapper/gradle-wrapper.properties: -------------------------------------------------------------------------------- 1 | #Sun Mar 11 14:01:32 EAT 2018 2 | distributionBase=GRADLE_USER_HOME 3 | distributionPath=wrapper/dists 4 | zipStoreBase=GRADLE_USER_HOME 5 | zipStorePath=wrapper/dists 6 | distributionUrl=https\://services.gradle.org/distributions/gradle-3.4.1-bin.zip 7 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | *.class 2 | *.log 3 | 4 | .idea 5 | build/ 6 | out/ 7 | .gradle 8 | .cache-* 9 | 10 | 11 | # sbt specific 12 | .cache 13 | .history 14 | .lib/ 15 | dist/* 16 | target/ 17 | lib_managed/ 18 | src_managed/ 19 | project/boot/ 20 | project/plugins/project/ 21 | 22 | # Scala-IDE specific 23 | .scala_dependencies 24 | .worksheet 25 | 26 | #needed for hiveContext 27 | metastore_db 28 | -------------------------------------------------------------------------------- /src/main/scala/ru/ispras/pu4spark/ProbabilisticClassifierConfig.scala: -------------------------------------------------------------------------------- 1 | package ru.ispras.pu4spark 2 | 3 | import org.apache.spark.ml.classification._ 4 | import org.apache.spark.mllib.linalg.Vector 5 | 6 | /** 7 | * @author Nikita Astrakhantsev (astrakhantsev@ispras.ru) 8 | */ 9 | sealed trait ProbabilisticClassifierConfig 10 | 11 | case class LogisticRegressionConfig(maxIter: Int = 100, 12 | regParam: Double = 1.0e-8, 13 | elasticNetParam: Double = 0.0) 14 | extends ProbabilisticClassifierConfig { 15 | def build(): ProbabilisticClassifier[Vector, LogisticRegression, LogisticRegressionModel] = { 16 | new LogisticRegression() 17 | .setLabelCol(ProbabilisticClassifierConfig.labelName).setFeaturesCol(ProbabilisticClassifierConfig.featuresName) 18 | .setMaxIter(maxIter).setRegParam(regParam).setElasticNetParam(elasticNetParam) 19 | } 20 | } 21 | 22 | case class RandomForestConfig(numTrees: Int = 512) 23 | extends ProbabilisticClassifierConfig { 24 | def build(): ProbabilisticClassifier[Vector, RandomForestClassifier, RandomForestClassificationModel] = { 25 | new RandomForestClassifier() 26 | .setLabelCol(ProbabilisticClassifierConfig.labelName).setFeaturesCol(ProbabilisticClassifierConfig.featuresName) 27 | .setNumTrees(numTrees) 28 | } 29 | } 30 | 31 | object ProbabilisticClassifierConfig { 32 | val labelName = "label" 33 | val featuresName = "indexedFeatures" 34 | val subclasses = List(classOf[LogisticRegressionConfig], classOf[RandomForestConfig]) 35 | } 36 | -------------------------------------------------------------------------------- /src/main/scala/ru/ispras/pu4spark/PositiveUnlabeledLearner.scala: -------------------------------------------------------------------------------- 1 | package ru.ispras.pu4spark 2 | 3 | import org.apache.spark.sql.DataFrame 4 | 5 | /** 6 | * Performs positive unlabeled (PU) learning, i.e. training a binary classifier in a semi-supervised way 7 | * from only positive and unlabeled examples 8 | * 9 | * @author Nikita Astrakhantsev (astrakhantsev@ispras.ru) 10 | */ 11 | trait PositiveUnlabeledLearner { 12 | 13 | /** 14 | * Updates dataframe by applying positive-unlabeled learning (append column with result of classification). 15 | * 16 | * @param df dataframe containing, among others, column with labels and features to be used in PU-learning 17 | * @param labelColumnName name for column containing 1 - positives and 0 - unlabeled marks for each instance 18 | * @param featuresColumnName name for 1 column containing features array (e.g. after VectorAssembler) 19 | * @param finalLabel name for column containing labels of final classification (1 for positive and -1 for negatives) 20 | * @return dataframe with new column corresponding to final classification 21 | */ 22 | def weight(df: DataFrame, 23 | labelColumnName: String = "featuresCol", 24 | featuresColumnName: String = "labelCol", 25 | finalLabel: String = "finalLabel"): DataFrame 26 | } 27 | 28 | /** 29 | * Subclasses should be case classes in order to be easily serializable (e.g. to JSON) 30 | */ 31 | trait PositiveUnlabeledLearnerConfig { 32 | def build(): PositiveUnlabeledLearner 33 | } 34 | 35 | /** 36 | * Needed for serialization by json4s (should be passed to org.json4s.ShortTypeHints) 37 | */ 38 | object PositiveUnlabeledLearnerConfig { 39 | val subclasses = List(classOf[TraditionalPULearnerConfig], classOf[GradualReductionPULearnerConfig]) 40 | } 41 | -------------------------------------------------------------------------------- /gradlew.bat: -------------------------------------------------------------------------------- 1 | @if "%DEBUG%" == "" @echo off 2 | @rem ########################################################################## 3 | @rem 4 | @rem Gradle startup script for Windows 5 | @rem 6 | @rem ########################################################################## 7 | 8 | @rem Set local scope for the variables with windows NT shell 9 | if "%OS%"=="Windows_NT" setlocal 10 | 11 | @rem Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script. 12 | set DEFAULT_JVM_OPTS= 13 | 14 | set DIRNAME=%~dp0 15 | if "%DIRNAME%" == "" set DIRNAME=. 16 | set APP_BASE_NAME=%~n0 17 | set APP_HOME=%DIRNAME% 18 | 19 | @rem Find java.exe 20 | if defined JAVA_HOME goto findJavaFromJavaHome 21 | 22 | set JAVA_EXE=java.exe 23 | %JAVA_EXE% -version >NUL 2>&1 24 | if "%ERRORLEVEL%" == "0" goto init 25 | 26 | echo. 27 | echo ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH. 28 | echo. 29 | echo Please set the JAVA_HOME variable in your environment to match the 30 | echo location of your Java installation. 31 | 32 | goto fail 33 | 34 | :findJavaFromJavaHome 35 | set JAVA_HOME=%JAVA_HOME:"=% 36 | set JAVA_EXE=%JAVA_HOME%/bin/java.exe 37 | 38 | if exist "%JAVA_EXE%" goto init 39 | 40 | echo. 41 | echo ERROR: JAVA_HOME is set to an invalid directory: %JAVA_HOME% 42 | echo. 43 | echo Please set the JAVA_HOME variable in your environment to match the 44 | echo location of your Java installation. 45 | 46 | goto fail 47 | 48 | :init 49 | @rem Get command-line arguments, handling Windowz variants 50 | 51 | if not "%OS%" == "Windows_NT" goto win9xME_args 52 | if "%@eval[2+2]" == "4" goto 4NT_args 53 | 54 | :win9xME_args 55 | @rem Slurp the command line arguments. 56 | set CMD_LINE_ARGS= 57 | set _SKIP=2 58 | 59 | :win9xME_args_slurp 60 | if "x%~1" == "x" goto execute 61 | 62 | set CMD_LINE_ARGS=%* 63 | goto execute 64 | 65 | :4NT_args 66 | @rem Get arguments from the 4NT Shell from JP Software 67 | set CMD_LINE_ARGS=%$ 68 | 69 | :execute 70 | @rem Setup the command line 71 | 72 | set CLASSPATH=%APP_HOME%\gradle\wrapper\gradle-wrapper.jar 73 | 74 | @rem Execute Gradle 75 | "%JAVA_EXE%" %DEFAULT_JVM_OPTS% %JAVA_OPTS% %GRADLE_OPTS% "-Dorg.gradle.appname=%APP_BASE_NAME%" -classpath "%CLASSPATH%" org.gradle.wrapper.GradleWrapperMain %CMD_LINE_ARGS% 76 | 77 | :end 78 | @rem End local scope for the variables with windows NT shell 79 | if "%ERRORLEVEL%"=="0" goto mainEnd 80 | 81 | :fail 82 | rem Set variable GRADLE_EXIT_CONSOLE if you need the _script_ return code instead of 83 | rem the _cmd.exe /c_ return code! 84 | if not "" == "%GRADLE_EXIT_CONSOLE%" exit 1 85 | exit /b 1 86 | 87 | :mainEnd 88 | if "%OS%"=="Windows_NT" endlocal 89 | 90 | :omega 91 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # pu4spark 2 | A library for [Positive-Unlabeled Learning](https://en.wikipedia.org/wiki/One-class_classification#PU_learning) 3 | for Apache Spark MLlib (ml package) 4 | 5 | ## Implemented algorithms 6 | 7 | ### Traditional PU 8 | Original Positive-Unlabeled learning algorithm; firstly proposed in 9 | > Liu, B., Dai, Y., Li, X. L., Lee, W. S., & Philip, Y. (2002). 10 | Partially supervised classification of text documents. 11 | In ICML 2002, Proceedings of the nineteenth international conference on machine learning. (pp. 387–394). 12 | 13 | ### Gradual Reduction PU (aka PU-LEA) 14 | Modified Positive-Unlabeled learning algorithm; 15 | main idea is to gradually refine set of positive examples. 16 | Pseudocode was taken from: 17 | >Fusilier, D. H., Montes-y-Gómez, M., Rosso, P., & Cabrera, R. G. (2015). 18 | Detecting positive and negative deceptive opinions using PU-learning. 19 | Information Processing & Management, 51(4), 433-443. 20 | 21 | ## Requirements 22 | 23 | Spark 1.5+ 24 | 25 | (Spark 2+ was not tested, 26 | but should work if replace `SparkContext` by `SparkSession` 27 | and `mllib.linalg.Vector` by `ml.linalg.Vector`) 28 | 29 | ## Linking 30 | 31 | The library is published into Maven central and JCenter. 32 | Add the following lines depending on your build system. 33 | 34 | ### Gradle 35 | 36 | ```gradle 37 | compile 'ru.ispras:pu4spark:0.3' 38 | ``` 39 | 40 | ### Maven 41 | 42 | ```xml 43 | 44 | ru.ispras 45 | pu4spark 46 | 0.3 47 | 48 | ``` 49 | 50 | ### SBT 51 | 52 | ``` 53 | libraryDependencies += "ru.ispras" % "pu4spark" % "0.3" 54 | ``` 55 | 56 | ## Building from Sources 57 | 58 | Build library with gradle: 59 | 60 | ```shell 61 | ./gradlew jar 62 | ``` 63 | 64 | ## Usage example 65 | 66 | 67 | ```scala 68 | val inputLabelName = "category" 69 | val srcFeaturesName = "srcFeatures" 70 | val outputLabel = "outputLabel" 71 | 72 | val puLearnerConfig = TraditionalPULearnerConfig(0.05, 1, LogisticRegressionConfig()) 73 | val puLearner = puLearnerConfig.build() 74 | val df = ... //needed df that contains at least the following columns: 75 | // binary label for positive and unlabel (inputLabelName) 76 | // and features assembled as vector (featuresName) 77 | 78 | val weightedDF = puLearner.weight(preparedDf, inputLabelName, srcFeaturesName, outputLabel) 79 | ``` 80 | Returned dataframe contains probability estimation for each instance in the column `outputLabel`. 81 | 82 | Features can be assembled to one column by using [VectorAssembler](https://spark.apache.org/docs/1.6.2/ml-features.html#vectorassembler): 83 | ```scala 84 | val assembler = new VectorAssembler() 85 | .setInputCols(df.columns.filter(c => c != rowName)) //keep here only feature columns 86 | .setOutputCol(featuresName) 87 | val pipeline = new Pipeline().setStages(Array(assembler)) 88 | val preparedDf = pipeline.fit(df).transform(df) 89 | ``` 90 | -------------------------------------------------------------------------------- /src/main/scala/ru/ispras/pu4spark/TraditionalPULearner.scala: -------------------------------------------------------------------------------- 1 | package ru.ispras.pu4spark 2 | 3 | import org.apache.logging.log4j.LogManager 4 | import org.apache.spark.ml.classification.{LogisticRegressionModel, ProbabilisticClassificationModel, ProbabilisticClassifier} 5 | import org.apache.spark.mllib.linalg.Vector 6 | import org.apache.spark.sql.DataFrame 7 | import org.apache.spark.sql.functions._ 8 | 9 | /** 10 | * Original Positive-Unlabeled learning algorithm; firstly proposed in 11 | * Liu, B., Dai, Y., Li, X. L., Lee, W. S., & Philip, Y. (2002). 12 | * Partially supervised classification of text documents. 13 | * In ICML 2002, Proceedings of the nineteenth international conference on machine learning. (pp. 387–394).
14 | * 15 | * Pseudocode was taken from: 16 | * Fusilier, D. H., Montes-y-Gómez, M., Rosso, P., & Cabrera, R. G. (2015). 17 | * Detecting positive and negative deceptive opinions using PU-learning. 18 | * Information Processing & Management, 51(4), 433-443. 19 | * 20 | * @author Nikita Astrakhantsev (astrakhantsev@ispras.ru) 21 | */ 22 | class TraditionalPULearner[ 23 | E <: ProbabilisticClassifier[Vector, E, M], 24 | M <: ProbabilisticClassificationModel[Vector, M]]( 25 | relNegThreshold: Double, 26 | maxIters: Int, 27 | classifier: ProbabilisticClassifier[Vector, E, M]) extends TwoStepPULearner[E,M](classifier) { 28 | val log = LogManager.getLogger(getClass) 29 | 30 | override def weight(df: DataFrame, labelColumnName: String, featuresColumnName: String, finalLabel: String): DataFrame = { 31 | val oneStepPUDF: DataFrame = zeroStep(df, labelColumnName, featuresColumnName, finalLabel) 32 | .drop("probability").drop("prediction").drop("rawPrediction").drop(ProbabilisticClassifierConfig.labelName) 33 | 34 | val confAdder = new RelNegConfidenceThresholdAdder(relNegThreshold) 35 | 36 | val prevLabel = "prevLabel" 37 | val curLabel = "curLabel" 38 | 39 | // replace all zeros with labels for undefined 40 | var curDF = replaceZerosByUndefLabel(oneStepPUDF, labelColumnName, prevLabel, TraditionalPULearner.undefLabel) 41 | 42 | for (i <- 1 to maxIters) { 43 | //replace weights by binary column for further learning (induce labels for curLabDF) 44 | val curLabelColumn = confAdder.binarizeUDF(curDF(finalLabel), curDF(prevLabel)) 45 | 46 | curDF = curDF.withColumn(curLabel, curLabelColumn).cache() 47 | val newRelNegCount = curDF 48 | //unlabeled in previous iterations && negative in current iteration 49 | .filter(curDF(prevLabel) === TraditionalPULearner.undefLabel && curDF(curLabel) === TraditionalPULearner.relNegLabel) 50 | .count() 51 | 52 | log.debug(s"newRelNegCount: $newRelNegCount") 53 | if (newRelNegCount == 0) { 54 | return curDF 55 | } 56 | 57 | //learn new classifier 58 | val curLabDF = curDF.filter(curDF(curLabel) !== TraditionalPULearner.undefLabel) //keep only positives and relnegs 59 | 60 | val newPreparedDf = indexLabelColumn(curLabDF, curLabel, ProbabilisticClassifierConfig.labelName, 61 | Seq(TraditionalPULearner.relNegLabel.toString, "1.0")) 62 | 63 | val model = classifier.fit(newPreparedDf) 64 | 65 | // log.debug(s"Coefficients: ${model.asInstanceOf[LogisticRegressionModel].coefficients} " + 66 | // s"Intercept: ${model.asInstanceOf[LogisticRegressionModel].intercept}") 67 | 68 | //apply classifier to still unlabeled data 69 | val labUnlabDF = model.transform(curDF) 70 | curDF = labUnlabDF.withColumn(finalLabel, getPOne(labUnlabDF("probability"))) 71 | .drop("probability").drop("prediction").drop("rawPrediction").drop(ProbabilisticClassifierConfig.labelName) 72 | curDF = curDF.drop(prevLabel) 73 | .withColumnRenamed(curLabel, prevLabel) 74 | } 75 | curDF 76 | } 77 | } 78 | 79 | private class RelNegConfidenceThresholdAdder(threshold: Double) extends Serializable { 80 | def binarize(probPred: Double, prevLabel: Int): Int = if (prevLabel == TraditionalPULearner.undefLabel) { // unlabeled 81 | if (probPred < threshold) { 82 | TraditionalPULearner.relNegLabel 83 | } else { 84 | TraditionalPULearner.undefLabel 85 | } 86 | } else { 87 | prevLabel // keep as it was (positive or reliable negatives, i.e. 1 or 0) 88 | } 89 | 90 | val binarizeUDF = udf(binarize(_: Double, _: Int)) 91 | } 92 | 93 | object TraditionalPULearner { 94 | val relNegLabel = 0 95 | val undefLabel = -1 96 | } 97 | 98 | case class TraditionalPULearnerConfig(relNegThreshold: Double = 0.5, 99 | maxIters: Int = 1, 100 | classifierConfig: ProbabilisticClassifierConfig = LogisticRegressionConfig() 101 | ) extends PositiveUnlabeledLearnerConfig { 102 | override def build(): PositiveUnlabeledLearner = { 103 | classifierConfig match { 104 | case lrc: LogisticRegressionConfig => new TraditionalPULearner(relNegThreshold, maxIters, lrc.build()) 105 | case rfc: RandomForestConfig => new TraditionalPULearner(relNegThreshold, maxIters, rfc.build()) 106 | } 107 | } 108 | } 109 | -------------------------------------------------------------------------------- /src/main/scala/ru/ispras/pu4spark/TwoStepPULearner.scala: -------------------------------------------------------------------------------- 1 | package ru.ispras.pu4spark 2 | 3 | import org.apache.spark.ml.Pipeline 4 | import org.apache.spark.ml.attribute.NominalAttribute 5 | import org.apache.spark.ml.classification.{ProbabilisticClassificationModel, ProbabilisticClassifier} 6 | import org.apache.spark.ml.feature.VectorIndexer 7 | import org.apache.spark.mllib.linalg.Vector 8 | import org.apache.spark.sql.DataFrame 9 | import org.apache.spark.sql.functions.{col, udf, when} 10 | import org.apache.spark.sql.types.DoubleType 11 | 12 | /** 13 | * Performs PU learning in a 2-step manner: 14 | * on the first step, choose between all unlabeled examples those are negative with high probability 15 | * (so called, reliable negatives), 16 | * so that at the second step use them along with positive examples for training binary classifier. 17 | * 18 | * @author Nikita Astrakhantsev (astrakhantsev@ispras.ru) 19 | */ 20 | abstract class TwoStepPULearner[ 21 | E <: ProbabilisticClassifier[Vector, E, M], 22 | M <: ProbabilisticClassificationModel[Vector, M]]( 23 | classifier: ProbabilisticClassifier[Vector, E, M]) extends PositiveUnlabeledLearner { 24 | 25 | /** 26 | * Extracts probability instead of binary prediction 27 | */ 28 | val getPOne = udf((v: Vector) => v(1)) 29 | 30 | /** 31 | * Train binary classifier by considering all unlabeled data as negative data, 32 | * then apply it to all unlabeled data in order to have some measure of reliability of these negatives. 33 | * 34 | * @param df dataframe to work with 35 | * @param labelColumnName name for column containing positive or unlabeled label 36 | * @param featuresColumnName name for column containing features as a vector (e.g. after VectorAssembler) 37 | * @param finalLabel name for column that will contain required measure of reliability of these negatives 38 | * @return updated dataframe with finalLabel column 39 | */ 40 | def zeroStep(df: DataFrame, labelColumnName: String, featuresColumnName: String, finalLabel: String): DataFrame = { 41 | val dfWithMeta = indexLabelColumn(df, labelColumnName, ProbabilisticClassifierConfig.labelName, Seq("0", "1")) 42 | 43 | //scaler seems to not improve results 44 | //StandardScaler with mean scaling requires DenseVectors, while VectorAssembler can return only SparseVectors 45 | // val scaler = new MinMaxScaler().setInputCol(srcFeaturesName).setOutputCol(scaledFeaturesName) 46 | 47 | // RF requires that (from Spark example, 'Automatically identify categorical features, and index them') 48 | val featureIndexer = new VectorIndexer() 49 | .setInputCol(featuresColumnName) 50 | .setOutputCol(ProbabilisticClassifierConfig.featuresName) 51 | .setMaxCategories(4) //features with > 4 distinct values are treated as continuous. 52 | 53 | val pipeline = new Pipeline().setStages(Array(featureIndexer)) 54 | val preparedDf = pipeline.fit(dfWithMeta).transform(dfWithMeta) 55 | 56 | val model: M = classifier.fit(preparedDf) 57 | val predictions: DataFrame = model.transform(preparedDf) 58 | val res = predictions.withColumn(finalLabel, getPOne(predictions("probability"))) 59 | res 60 | } 61 | 62 | /** 63 | * Adds meta-information to label column and casts it to DoubleType, so that it can be used for training. 64 | * StringIndexer can't be used, because it assigns index based on labels frequency. 65 | * 66 | * 67 | * @param df dataframe to index label 68 | * @param inputCol name of column with original label 69 | * @param outputCol name of column with indexed label 70 | * @param values labels to support 71 | * @return dataframe with indexed label 72 | */ 73 | def indexLabelColumn(df: DataFrame, inputCol: String, outputCol: String, values: Seq[String]): DataFrame = { 74 | val meta = NominalAttribute 75 | .defaultAttr 76 | .withName(inputCol) 77 | .withValues(values.head, values.tail: _*) 78 | .toMetadata 79 | 80 | df.withColumn(outputCol, col(inputCol).as(outputCol, meta).cast(DoubleType)) 81 | } 82 | 83 | /** 84 | * Replaces one value in column by another and renames this column. 85 | * It is used to change labels from zero to special value indicating undefined. 86 | * 87 | * @param df dataframe to replace value 88 | * @param origColName name of column with original label 89 | * @param newColName name of column with replaced label 90 | * @param value2replace value from that column that should be used instead of existing value 91 | * (i.e. if the value differs from value2keep, than it would be replaced by value2replace 92 | * @param value2keep value that should be kept 93 | * @return dataframe with replaced values 94 | */ 95 | def replaceZerosByUndefLabel(df: DataFrame, 96 | origColName: String, 97 | newColName: String, 98 | value2replace: Double, 99 | value2keep: Double = 1): DataFrame = { 100 | df.withColumn(newColName, 101 | when(col(origColName).equalTo(value2keep), value2keep).otherwise(value2replace)) 102 | .drop(origColName) 103 | } 104 | } 105 | -------------------------------------------------------------------------------- /gradlew: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env bash 2 | 3 | ############################################################################## 4 | ## 5 | ## Gradle start up script for UN*X 6 | ## 7 | ############################################################################## 8 | 9 | # Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script. 10 | DEFAULT_JVM_OPTS="" 11 | 12 | APP_NAME="Gradle" 13 | APP_BASE_NAME=`basename "$0"` 14 | 15 | # Use the maximum available, or set MAX_FD != -1 to use that value. 16 | MAX_FD="maximum" 17 | 18 | warn ( ) { 19 | echo "$*" 20 | } 21 | 22 | die ( ) { 23 | echo 24 | echo "$*" 25 | echo 26 | exit 1 27 | } 28 | 29 | # OS specific support (must be 'true' or 'false'). 30 | cygwin=false 31 | msys=false 32 | darwin=false 33 | case "`uname`" in 34 | CYGWIN* ) 35 | cygwin=true 36 | ;; 37 | Darwin* ) 38 | darwin=true 39 | ;; 40 | MINGW* ) 41 | msys=true 42 | ;; 43 | esac 44 | 45 | # Attempt to set APP_HOME 46 | # Resolve links: $0 may be a link 47 | PRG="$0" 48 | # Need this for relative symlinks. 49 | while [ -h "$PRG" ] ; do 50 | ls=`ls -ld "$PRG"` 51 | link=`expr "$ls" : '.*-> \(.*\)$'` 52 | if expr "$link" : '/.*' > /dev/null; then 53 | PRG="$link" 54 | else 55 | PRG=`dirname "$PRG"`"/$link" 56 | fi 57 | done 58 | SAVED="`pwd`" 59 | cd "`dirname \"$PRG\"`/" >/dev/null 60 | APP_HOME="`pwd -P`" 61 | cd "$SAVED" >/dev/null 62 | 63 | CLASSPATH=$APP_HOME/gradle/wrapper/gradle-wrapper.jar 64 | 65 | # Determine the Java command to use to start the JVM. 66 | if [ -n "$JAVA_HOME" ] ; then 67 | if [ -x "$JAVA_HOME/jre/sh/java" ] ; then 68 | # IBM's JDK on AIX uses strange locations for the executables 69 | JAVACMD="$JAVA_HOME/jre/sh/java" 70 | else 71 | JAVACMD="$JAVA_HOME/bin/java" 72 | fi 73 | if [ ! -x "$JAVACMD" ] ; then 74 | die "ERROR: JAVA_HOME is set to an invalid directory: $JAVA_HOME 75 | 76 | Please set the JAVA_HOME variable in your environment to match the 77 | location of your Java installation." 78 | fi 79 | else 80 | JAVACMD="java" 81 | which java >/dev/null 2>&1 || die "ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH. 82 | 83 | Please set the JAVA_HOME variable in your environment to match the 84 | location of your Java installation." 85 | fi 86 | 87 | # Increase the maximum file descriptors if we can. 88 | if [ "$cygwin" = "false" -a "$darwin" = "false" ] ; then 89 | MAX_FD_LIMIT=`ulimit -H -n` 90 | if [ $? -eq 0 ] ; then 91 | if [ "$MAX_FD" = "maximum" -o "$MAX_FD" = "max" ] ; then 92 | MAX_FD="$MAX_FD_LIMIT" 93 | fi 94 | ulimit -n $MAX_FD 95 | if [ $? -ne 0 ] ; then 96 | warn "Could not set maximum file descriptor limit: $MAX_FD" 97 | fi 98 | else 99 | warn "Could not query maximum file descriptor limit: $MAX_FD_LIMIT" 100 | fi 101 | fi 102 | 103 | # For Darwin, add options to specify how the application appears in the dock 104 | if $darwin; then 105 | GRADLE_OPTS="$GRADLE_OPTS \"-Xdock:name=$APP_NAME\" \"-Xdock:icon=$APP_HOME/media/gradle.icns\"" 106 | fi 107 | 108 | # For Cygwin, switch paths to Windows format before running java 109 | if $cygwin ; then 110 | APP_HOME=`cygpath --path --mixed "$APP_HOME"` 111 | CLASSPATH=`cygpath --path --mixed "$CLASSPATH"` 112 | JAVACMD=`cygpath --unix "$JAVACMD"` 113 | 114 | # We build the pattern for arguments to be converted via cygpath 115 | ROOTDIRSRAW=`find -L / -maxdepth 1 -mindepth 1 -type d 2>/dev/null` 116 | SEP="" 117 | for dir in $ROOTDIRSRAW ; do 118 | ROOTDIRS="$ROOTDIRS$SEP$dir" 119 | SEP="|" 120 | done 121 | OURCYGPATTERN="(^($ROOTDIRS))" 122 | # Add a user-defined pattern to the cygpath arguments 123 | if [ "$GRADLE_CYGPATTERN" != "" ] ; then 124 | OURCYGPATTERN="$OURCYGPATTERN|($GRADLE_CYGPATTERN)" 125 | fi 126 | # Now convert the arguments - kludge to limit ourselves to /bin/sh 127 | i=0 128 | for arg in "$@" ; do 129 | CHECK=`echo "$arg"|egrep -c "$OURCYGPATTERN" -` 130 | CHECK2=`echo "$arg"|egrep -c "^-"` ### Determine if an option 131 | 132 | if [ $CHECK -ne 0 ] && [ $CHECK2 -eq 0 ] ; then ### Added a condition 133 | eval `echo args$i`=`cygpath --path --ignore --mixed "$arg"` 134 | else 135 | eval `echo args$i`="\"$arg\"" 136 | fi 137 | i=$((i+1)) 138 | done 139 | case $i in 140 | (0) set -- ;; 141 | (1) set -- "$args0" ;; 142 | (2) set -- "$args0" "$args1" ;; 143 | (3) set -- "$args0" "$args1" "$args2" ;; 144 | (4) set -- "$args0" "$args1" "$args2" "$args3" ;; 145 | (5) set -- "$args0" "$args1" "$args2" "$args3" "$args4" ;; 146 | (6) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" ;; 147 | (7) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" "$args6" ;; 148 | (8) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" "$args6" "$args7" ;; 149 | (9) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" "$args6" "$args7" "$args8" ;; 150 | esac 151 | fi 152 | 153 | # Split up the JVM_OPTS And GRADLE_OPTS values into an array, following the shell quoting and substitution rules 154 | function splitJvmOpts() { 155 | JVM_OPTS=("$@") 156 | } 157 | eval splitJvmOpts $DEFAULT_JVM_OPTS $JAVA_OPTS $GRADLE_OPTS 158 | JVM_OPTS[${#JVM_OPTS[*]}]="-Dorg.gradle.appname=$APP_BASE_NAME" 159 | 160 | exec "$JAVACMD" "${JVM_OPTS[@]}" -classpath "$CLASSPATH" org.gradle.wrapper.GradleWrapperMain "$@" 161 | -------------------------------------------------------------------------------- /src/main/scala/ru/ispras/pu4spark/GradualReductionPULearner.scala: -------------------------------------------------------------------------------- 1 | package ru.ispras.pu4spark 2 | 3 | import org.apache.logging.log4j.LogManager 4 | import org.apache.spark.ml.classification.{ProbabilisticClassificationModel, ProbabilisticClassifier} 5 | import org.apache.spark.mllib.linalg.Vector 6 | import org.apache.spark.sql.DataFrame 7 | import org.apache.spark.sql.functions._ 8 | 9 | /** 10 | * Modified Positive-Unlabeled learning algorithm; main idea is to gradually refine set of positive examples
11 | * 12 | * Pseudocode was taken from: 13 | * Fusilier, D. H., Montes-y-Gómez, M., Rosso, P., & Cabrera, R. G. (2015). 14 | * Detecting positive and negative deceptive opinions using PU-learning. 15 | * Information Processing & Management, 51(4), 433-443. 16 | * 17 | * @author Nikita Astrakhantsev (astrakhantsev@ispras.ru) 18 | */ 19 | class GradualReductionPULearner[ 20 | E <: ProbabilisticClassifier[Vector, E, M], 21 | M <: ProbabilisticClassificationModel[Vector, M]]( 22 | relNegThreshold: Double, 23 | classifier: ProbabilisticClassifier[Vector, E, M]) extends TwoStepPULearner[E,M](classifier) { 24 | 25 | val log = LogManager.getLogger(getClass) 26 | 27 | override def weight(df: DataFrame, labelColumnName: String, featuresColumnName: String, finalLabel: String): DataFrame = { 28 | val oneStepPUDF: DataFrame = zeroStep(df, labelColumnName, featuresColumnName, finalLabel) 29 | .drop("probability").drop("prediction").drop("rawPrediction").drop(ProbabilisticClassifierConfig.labelName) 30 | 31 | val prevLabel = "prevLabel" 32 | val curLabel = "curLabel" 33 | var curDF = replaceZerosByUndefLabel(oneStepPUDF, labelColumnName, prevLabel, GradualReductionPULearner.undefLabel) 34 | 35 | val confAdder = new GradRelNegConfidenceThresholdAdder(relNegThreshold, GradualReductionPULearner.undefLabel) 36 | 37 | //replace weights by binary column for further learning (induce labels for curLabDF) 38 | val curLabelColumn = confAdder.binarizeUDF(curDF(finalLabel), curDF(prevLabel)) 39 | 40 | curDF = curDF.withColumn(curLabel, curLabelColumn).cache() 41 | var newRelNegCount = curDF 42 | //unlabeled in previous iterations && negative in current iteration 43 | .filter(curDF(prevLabel) === GradualReductionPULearner.undefLabel && curDF(curLabel) === GradualReductionPULearner.relNegLabel) 44 | .count() 45 | 46 | log.debug(s"newRelNegCount: $newRelNegCount") 47 | var prevNewRelNegCount = newRelNegCount 48 | val totalPosCount = curDF.filter(curDF(curLabel) === GradualReductionPULearner.posLabel).count() 49 | var totalRelNegCount = curDF.filter(curDF(curLabel) === GradualReductionPULearner.relNegLabel).count() 50 | 51 | var prevGain = Long.MaxValue 52 | var curGain = newRelNegCount 53 | 54 | do { 55 | //learn new classifier 56 | val curLabDF = curDF.filter(curDF(curLabel) !== GradualReductionPULearner.undefLabel) 57 | 58 | val newPreparedDf = indexLabelColumn(curLabDF, curLabel, ProbabilisticClassifierConfig.labelName, 59 | Seq(GradualReductionPULearner.relNegLabel.toString, GradualReductionPULearner.posLabel.toString)) 60 | 61 | val model = classifier.fit(newPreparedDf) 62 | 63 | //apply classifier to all data (however, we are interested in ReliableNegatives data only, see confAdder) 64 | val labUnlabDF = model.transform(curDF) 65 | curDF = labUnlabDF.withColumn(finalLabel, getPOne(labUnlabDF("probability"))) 66 | .drop("probability").drop("prediction").drop("rawPrediction").drop(ProbabilisticClassifierConfig.labelName) 67 | curDF = curDF.drop(prevLabel) 68 | .withColumnRenamed(curLabel, prevLabel) 69 | 70 | val innerConfAdder = new GradRelNegConfidenceThresholdAdder(relNegThreshold, GradualReductionPULearner.relNegLabel) 71 | val curLabelColumn = innerConfAdder.binarizeUDF(curDF(finalLabel), curDF(prevLabel)) 72 | 73 | curDF = curDF.withColumn(curLabel, curLabelColumn).cache() 74 | prevNewRelNegCount = newRelNegCount 75 | newRelNegCount = curDF 76 | //negative in current iteration 77 | .filter(curDF(curLabel) === GradualReductionPULearner.relNegLabel) 78 | .count() 79 | totalRelNegCount = curDF.filter(curDF(curLabel) === GradualReductionPULearner.relNegLabel).count() 80 | prevGain = curGain 81 | curGain = prevNewRelNegCount - totalRelNegCount 82 | log.debug(s"newRelNegCount: $newRelNegCount, prevNewRelNegCount: $prevNewRelNegCount, totalRelNegCount: $totalRelNegCount") 83 | log.debug(s"curGain: $curGain, prevGain: $prevGain") 84 | } while (curGain > 0 && curGain < prevGain && totalPosCount < totalRelNegCount) 85 | curDF 86 | } 87 | } 88 | 89 | private class GradRelNegConfidenceThresholdAdder(threshold: Double, labelToConsider: Int) extends Serializable { 90 | def binarize(probPred: Double, prevLabel: Int): Int = if (prevLabel == labelToConsider) { 91 | if (probPred < threshold) { 92 | GradualReductionPULearner.relNegLabel 93 | } else { 94 | GradualReductionPULearner.undefLabel 95 | } 96 | } else { 97 | prevLabel // keep as it was //(1 or -1 in case of unlabeled classification) 98 | } 99 | 100 | val binarizeUDF = udf(binarize(_: Double, _: Int)) 101 | } 102 | 103 | object GradualReductionPULearner { 104 | val relNegLabel = 0 105 | val posLabel = 1 106 | val undefLabel = -1 107 | } 108 | 109 | case class GradualReductionPULearnerConfig(relNegThreshold: Double = 0.5, 110 | classifierConfig: ProbabilisticClassifierConfig) extends PositiveUnlabeledLearnerConfig { 111 | override def build(): PositiveUnlabeledLearner = { 112 | classifierConfig match { 113 | case lrc: LogisticRegressionConfig => new GradualReductionPULearner(relNegThreshold, lrc.build()) 114 | case rfc: RandomForestConfig => new GradualReductionPULearner(relNegThreshold, rfc.build()) 115 | } 116 | } 117 | } 118 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "{}" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright {yyyy} {name of copyright owner} 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | --------------------------------------------------------------------------------