基于名称(标签)的Scala随机森林特征重要性提取

基于名称(标签)的Scala随机森林特征重要性提取,scala,apache-spark,machine-learning,random-forest,feature-selection,Scala,Apache Spark,Machine Learning,Random Forest,Feature Selection,有没有办法从模型中提取特征重要性并附加featureCols名称以便于分析 我有点像: val featureCols = Array("a","b","c".......... like 67 more) val assembler = new VectorAssembler().setInputCols(featureCols).setOutputCol("features") val df2 = assembler.transform(modeling_db) val labelInde

有没有办法从模型中提取特征重要性并附加
featureCols
名称以便于分析

我有点像:

val featureCols = Array("a","b","c".......... like 67 more)

val assembler = new VectorAssembler().setInputCols(featureCols).setOutputCol("features")
val df2 = assembler.transform(modeling_db)
val labelIndexer = new StringIndexer().setInputCol("def").setOutputCol("label")
val df3 = labelIndexer.fit(df2).transform(df2)
val splitSeed = 5043
val Array(trainingData, testDataCE) = df3.randomSplit(Array(0.7, 0.3), splitSeed)
val classifier = new RandomForestClassifier().setImpurity("gini").setMaxDepth(19).setNumTrees(57).setFeatureSubsetStrategy("auto").setSeed(5043)
val model = classifier.fit(trainingData)
之后,我们尝试通过以下方式提取重要性:

model.featureImportances
答案很难分析:

res14: org.apache.spark.mllib.linalg.Vector = (71,[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,20,23,25,27,33,34,35,38,39,41,42,45,47,48,49,50,51,52,53,54,55,56,57,58,60,61,62,63,64,65,66,67,68,69,70],[0.22362951804309808,0.1830148359365108,0.10246542303449771,0.1699399958851977,0.06486419413350401,0.05187244974385025,0.02627047699833213,0.014498050071723645,0.026182513062665076,0.007126662761055224,0.012375060477018274,0.004354513006816487,0.004361008357237427,0.008435852744278544,0.003195472326415685,0.0023071401643885753,0.004602370417578224,0.0030394399903992345,6.92408316823549E-4,0.011207695216651398,7.609910745572573E-4,8.316382113306638E-4,0.0021506289318167916,0.0013468620354363688,0.006968754359778437,0.018796331618729723,0.0024516591941419444,0.005980997035580654,0.0027983...

有没有一种方法可以将此答案“upack”并附加到原始标签名称中?

您在
featureCols
中有原始列名,并且似乎没有涉及任何向量,因此您可以简单地将这两个数组组合在一起。对于这样的输入数据:

val featureCols = Array("a", "b", "c", "d", "e")
val featureImportance = Vectors.dense(Array(0.15, 0.25, 0.1, 0.35, 0.15)).toSparse
干脆

val res = featureCols.zip(featureImportance.toArray).sortBy(-_._2)
通过打印将导致

(d,0.35)
(b,0.25)
(a,0.15)
(e,0.15)
(c,0.1)