关于spark mllib中 FPM算法的性能问题

[qmd1178920084]

鄙人刚开始接触spark大数据处理，最近在做一些数据挖掘的实验，其中一个对比算法借用了spark库中的经典数据挖掘算法FP-Growth，我这里的集群是三台机器：1*master(2cores，4G mem)， 2*worker（4cores，8G mem）。在进行的实验中挖掘的数据集是利用IBM的数据产生器生成的T40I10D100K.txt，大小14.76 MB。因为该数据集中重复出现的数字较少，所以支持度设为了1%。分区数为16。结果挖掘过程中，出现了堆溢出的错误，应该是由于growth过程中递归建立条件FP-Tree。这里我就产生了一个疑问，我的集群两台worker加起来可用内存也有12GB的样子，怎么会连15MB的数据都处理不了。是因为支持度过低，还是因为FP-Growth进行频繁项集的挖掘势必会造成这么大的内存消耗？

请各位有经验的前辈赐教，谢谢。

[qmd1178920084]

这是调用FP-Growth的源码

object PFP {
  def main(args: Array[String]): Unit = {
    val texts = mutable.Map(
   // "T25I10D10K.txt"->List(0.005,0.004,0.003,0.002),
   // "mushroom.txt"->List(0.01))
   // "chess.txt"->List(0.4))
   // "accidents.txt"->List(0.1))
   // "T10I4D100K.txt"->List(0.005,0.004,0.003,0.002,0.001),
    //"T40I10D100K.txt"->List(0.01))
    "connect-4.txt"->List(0.3))
   // "kddcup99.txt"->List(0.0001,0.00009,0.00008,0.00007,0.00006))
   //"USCensus.txt"->List(0.5))
 //  "connect-4.txt"->List(0.5))

    val conf =new SparkConf().setAppName("PFP_scala")
    val sc =new SparkContext(conf)

    texts.foreach{ text =>
    val writer =new PrintWriter(new File("/root/app/scala2.10/PFP/"+text._1))
    val data= sc.textFile("/usr/local/eclipsews/"+text._1)
    val startTime = System.currentTimeMillis()
    val transactions: RDD[Array[String]] = data.map(s => s.trim.split(' '))
    val ioTime = System.currentTimeMillis() - startTime

    text._2.foreach{ support =>
    for(i<-0 to 0){

    val time1 = System.currentTimeMillis()
    val fpg = new FPGrowth()
    .setMinSupport(support)
    .setNumPartitions(16)

    val model = fpg.run(transactions)
    val process = java.lang.Runtime.getRuntime.exec("/root/app/scala2.10/PFP/checkHDFS.sh")
    process.waitFor();
    model.freqItemsets.saveAsTextFile("/usr/local/PFP")

    val endTime = System.currentTimeMillis()
    val mineTime =endTime - time1
    //hehe.foreach { itemset =>println(itemset.items.mkString("[", ",", "]") + ", " + itemset.freq)}
    val time = mineTime + ioTime
    writer.write("database:  "+text._1+"  support:  "+support+ " iotime:  " +ioTime +" mineTime "+ mineTime+  "  time:  "+ time+ "\n")
      }
    }
    writer.close()
   }