在機(jī)器學(xué)習(xí)中,一般都會(huì)按照下面幾個(gè)步驟:特征提取、數(shù)據(jù)預(yù)處理、特征選擇、模型訓(xùn)練、檢驗(yàn)優(yōu)化。那么特征的選擇就很關(guān)鍵了,一般模型最后效果的好壞往往都是跟特征的選擇有關(guān)系的,因?yàn)槟P捅旧淼膮?shù)并沒(méi)有太多優(yōu)化的點(diǎn),反而特征這邊有時(shí)候多加一個(gè)或者少加一個(gè),最終的結(jié)果都會(huì)差別很大。
在SparkMLlib中為我們提供了幾種特征選擇的方法,分別是VectorSlicer
、RFormula
和ChiSqSelector
。
下面就介紹下這三個(gè)方法的使用,強(qiáng)烈推薦有時(shí)間的把參考的文獻(xiàn)都閱讀下,會(huì)有所收獲!
VectorSlicer
這個(gè)轉(zhuǎn)換器可以支持用戶自定義選擇列,可以基于下標(biāo)索引,也可以基于列名。
如果是下標(biāo)都可以使用setIndices方法
如果是列名可以使用setNames方法。使用這個(gè)方法的時(shí)候,vector字段需要通過(guò)AttributeGroup設(shè)置每個(gè)向量元素的列名。
注意1:可以同時(shí)使用setInices和setName
object VectorSlicer { def main(args: Array[String]) { val conf = new SparkConf().setAppName("VectorSlicer-Test").setMaster("local[2]") val sc = new SparkContext(conf) sc.setLogLevel("WARN") var sqlContext = new SQLContext(sc) val data = Array(Row(Vectors.dense(-2.0, 2.3, 0.0,1.0,2.0))) val defaultAttr = NumericAttribute.defaultAttr val attrs = Array("f1", "f2", "f3","f4","f5").map(defaultAttr.withName) val attrGroup = new AttributeGroup("userFeatures", attrs.asInstanceOf[Array[Attribute]]) val dataRDD = sc.parallelize(data) val dataset = sqlContext.createDataFrame(dataRDD, StructType(Array(attrGroup.toStructField()))) val slicer = new VectorSlicer().setInputCol("userFeatures").setOutputCol("features") slicer.setIndices(Array(0)).setNames(Array("f2")) val output = slicer.transform(dataset) println(output.select("userFeatures", "features").first()) } }
注意2:如果下標(biāo)和索引重復(fù),會(huì)報(bào)重復(fù)的錯(cuò):
比如:
slicer.setIndices(Array(1)).setNames(Array("f2"))
那么會(huì)遇到報(bào)錯(cuò)
Exception in thread "main" java.lang.IllegalArgumentException: requirement failed: VectorSlicer requires indices and names to be disjoint sets of features, but they overlap. indices: [1]. names: [1:f2] at scala.Predef$.require(Predef.scala:233) at org.apache.spark.ml.feature.VectorSlicer.getSelectedFeatureIndices(VectorSlicer.scala:137) at org.apache.spark.ml.feature.VectorSlicer.transform(VectorSlicer.scala:108) at xingoo.mllib.VectorSlicer$.main(VectorSlicer.scala:35) at xingoo.mllib.VectorSlicer.main(VectorSlicer.scala) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at com.intellij.rt.execution.application.AppMain.main(AppMain.java:144)
注意3:如果下標(biāo)不存在
slicer.setIndices(Array(6))
如果數(shù)組越界也會(huì)報(bào)錯(cuò)
Exception in thread "main" java.lang.ArrayIndexOutOfBoundsException: 6 at org.apache.spark.ml.feature.VectorSlicer$$anonfun$3$$anonfun$apply$2.apply(VectorSlicer.scala:110) at org.apache.spark.ml.feature.VectorSlicer$$anonfun$3$$anonfun$apply$2.apply(VectorSlicer.scala:110) at scala.collection.TraversableLike$$anonfun$map$1.apply(TraversableLike.scala:244) at scala.collection.TraversableLike$$anonfun$map$1.apply(TraversableLike.scala:244) at scala.collection.IndexedSeqOptimized$class.foreach(IndexedSeqOptimized.scala:33) at scala.collection.mutable.ArrayOps$ofInt.foreach(ArrayOps.scala:156) at scala.collection.TraversableLike$class.map(TraversableLike.scala:244) at scala.collection.mutable.ArrayOps$ofInt.map(ArrayOps.scala:156) at org.apache.spark.ml.feature.VectorSlicer$$anonfun$3.apply(VectorSlicer.scala:110) at org.apache.spark.ml.feature.VectorSlicer$$anonfun$3.apply(VectorSlicer.scala:109) at scala.Option.map(Option.scala:145) at org.apache.spark.ml.feature.VectorSlicer.transform(VectorSlicer.scala:109) at xingoo.mllib.VectorSlicer$.main(VectorSlicer.scala:35) at xingoo.mllib.VectorSlicer.main(VectorSlicer.scala) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at com.intellij.rt.execution.application.AppMain.main(AppMain.java:144)
注意4:如果名稱不存在也會(huì)報(bào)錯(cuò)
Exception in thread "main" java.lang.IllegalArgumentException: requirement failed: getFeatureIndicesFromNames found no feature with name f8 in column StructField(userFeatures,org.apache.spark.mllib.linalg.VectorUDT@f71b0bce,false). at scala.Predef$.require(Predef.scala:233) at org.apache.spark.ml.util.MetadataUtils$$anonfun$getFeatureIndicesFromNames$2.apply(MetadataUtils.scala:89) at org.apache.spark.ml.util.MetadataUtils$$anonfun$getFeatureIndicesFromNames$2.apply(MetadataUtils.scala:88) at scala.collection.TraversableLike$$anonfun$map$1.apply(TraversableLike.scala:244) at scala.collection.TraversableLike$$anonfun$map$1.apply(TraversableLike.scala:244) at scala.collection.IndexedSeqOptimized$class.foreach(IndexedSeqOptimized.scala:33) at scala.collection.mutable.ArrayOps$ofRef.foreach(ArrayOps.scala:108) at scala.collection.TraversableLike$class.map(TraversableLike.scala:244) at scala.collection.mutable.ArrayOps$ofRef.map(ArrayOps.scala:108) at org.apache.spark.ml.util.MetadataUtils$.getFeatureIndicesFromNames(MetadataUtils.scala:88) at org.apache.spark.ml.feature.VectorSlicer.getSelectedFeatureIndices(VectorSlicer.scala:129) at org.apache.spark.ml.feature.VectorSlicer.transform(VectorSlicer.scala:108) at xingoo.mllib.VectorSlicer$.main(VectorSlicer.scala:35) at xingoo.mllib.VectorSlicer.main(VectorSlicer.scala) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at com.intellij.rt.execution.application.AppMain.main(AppMain.java:144)
注意5:經(jīng)過(guò)特征選擇后,特征的順序與索引和名稱的順序相同
RFormula
這個(gè)轉(zhuǎn)換器可以幫助基于R模型,自動(dòng)生成feature和label。比如說(shuō)最常用的線性回歸,在先用回歸中,我們需要把一些離散化的變量變成啞變量,即轉(zhuǎn)變成onehot編碼,使之?dāng)?shù)值化,這個(gè)我之前的文章也介紹過(guò),這里就不多說(shuō)了。
如果不是用這個(gè)RFormula,我們可能需要經(jīng)過(guò)幾個(gè)步驟:
StringIndex...OneHotEncoder...
而且每個(gè)特征都要經(jīng)過(guò)這樣的變換,非常繁瑣。有了RFormula,幾乎可以一鍵把所有的特征問(wèn)題解決。
id | coutry | hour | clicked |
---|---|---|---|
7 | US | 18 | 1.0 |
8 | CA | 12 | 0.0 |
9 | NZ | 15 | 0.0 |
然后我們只要寫一個(gè)類似這樣的公式clicked ~ country + hour + my_test
,就代表clicked
為label
,coutry、hour、my_test
是三個(gè)特征
比如下面的代碼:
object RFormulaTest { def main(args: Array[String]) { val conf = new SparkConf().setAppName("RFormula-Test").setMaster("local[2]") val sc = new SparkContext(conf) sc.setLogLevel("WARN") var sqlContext = new SQLContext(sc) val dataset = sqlContext.createDataFrame(Seq( (7, "US", 18, 1.0,"a"), (8, "CA", 12, 0.0,"b"), (9, "NZ", 15, 0.0,"a") )).toDF("id", "country", "hour", "clicked","my_test") val formula = new RFormula() .setFormula("clicked ~ country + hour + my_test") .setFeaturesCol("features") .setLabelCol("label") val output = formula.fit(dataset).transform(dataset) output.show() output.select("features", "label").show() } }
得到的結(jié)果
+---+-------+----+-------+-------+------------------+-----+ | id|country|hour|clicked|my_test| features|label| +---+-------+----+-------+-------+------------------+-----+ | 7| US| 18| 1.0| a|[0.0,0.0,18.0,1.0]| 1.0| | 8| CA| 12| 0.0| b|[1.0,0.0,12.0,0.0]| 0.0| | 9| NZ| 15| 0.0| a|[0.0,1.0,15.0,1.0]| 0.0| +---+-------+----+-------+-------+------------------+-----+ +------------------+-----+ | features|label| +------------------+-----+ |[0.0,0.0,18.0,1.0]| 1.0| |[1.0,0.0,12.0,0.0]| 0.0| |[0.0,1.0,15.0,1.0]| 0.0| +------------------+-----+
ChiSqSelector
這個(gè)選擇器支持基于卡方檢驗(yàn)的特征選擇,卡方檢驗(yàn)是一種計(jì)算變量獨(dú)立性的檢驗(yàn)手段。具體的可以參考維基百科,最終的結(jié)論就是卡方的值越大,就是我們?cè)较胍奶卣?。因此這個(gè)選擇器就可以理解為,再計(jì)算卡方的值,最后按照這個(gè)值排序,選擇我們想要的個(gè)數(shù)的特征。
代碼也很簡(jiǎn)單
object ChiSqSelectorTest { def main(args: Array[String]) { val conf = new SparkConf().setAppName("ChiSqSelector-Test").setMaster("local[2]") val sc = new SparkContext(conf) sc.setLogLevel("WARN") var sqlContext = new SQLContext(sc) val data = Seq( (7, Vectors.dense(0.0, 0.0, 18.0, 1.0), 1.0), (8, Vectors.dense(0.0, 1.0, 12.0, 0.0), 0.0), (9, Vectors.dense(1.0, 0.0, 15.0, 0.1), 0.0) ) val beanRDD = sc.parallelize(data).map(t3 => Bean(t3._1,t3._2,t3._3)) val df = sqlContext.createDataFrame(beanRDD) val selector = new ChiSqSelector() .setNumTopFeatures(2) .setFeaturesCol("features") .setLabelCol("clicked") .setOutputCol("selectedFeatures") val result = selector.fit(df).transform(df) result.show() } case class Bean(id:Double,features:org.apache.spark.mllib.linalg.Vector,clicked:Double){} }
這樣得到的結(jié)果:
+---+------------------+-------+----------------+ | id| features|clicked|selectedFeatures| +---+------------------+-------+----------------+ |7.0|[0.0,0.0,18.0,1.0]| 1.0| [18.0,1.0]| |8.0|[0.0,1.0,12.0,0.0]| 0.0| [12.0,0.0]| |9.0|[1.0,0.0,15.0,0.1]| 0.0| [15.0,0.1]| +---+------------------+-------+----------------+
總結(jié)
下面總結(jié)一下三種特征選擇的使用場(chǎng)景:
VectorSilcer
,這個(gè)選擇器適合那種有很多特征,并且明確知道自己想要哪個(gè)特征的情況。比如你有一個(gè)很全的用戶畫像系統(tǒng),每個(gè)人有成百上千個(gè)特征,但是你指向抽取用戶對(duì)電影感興趣相關(guān)的特征,因此只要手動(dòng)選擇一下就可以了。RFormula
,這個(gè)選擇器適合在需要做OneHotEncoder的時(shí)候,可以一個(gè)簡(jiǎn)單的代碼把所有的離散特征轉(zhuǎn)化成數(shù)值化表示。ChiSqSelector
,卡方檢驗(yàn)選擇器適合在你有比較多的特征,但是不知道這些特征哪個(gè)有用,哪個(gè)沒(méi)用,想要通過(guò)某種方式幫助你快速篩選特征,那么這個(gè)方法很適合。
以上的總結(jié)純屬個(gè)人看法,不代表官方做法,如果有其他的見(jiàn)解可以留言~ 多交流!
http://www.cnblogs.com/xing901022/p/7152922.html