04_2 基于话题文本相似度的推荐_推荐

前面已经进行了分词。本文介绍如何进行推荐，核心代码如下：

 private val rdd: RDD[Topic_desc] = spark.sparkContext.textFile(SOURCE_URL).map {
    line =>
      val strings = line.split("\\$,")
      val idstring = strings(0).split("\\(")
      Topic_desc(idstring(1).toInt, strings(1))
  }
  private val pframe: DataFrame = rdd.toDF()
  //过滤掉空值
  pframe.registerTempTable("filter")
  private val filterFrame: DataFrame = spark.sqlContext.sql("select id,desc from filter where desc!=')'")
 //得到需要的文本
  private val dframerdd: RDD[Topic_desc_result] = filterFrame.rdd.map {
    line =>
      val str1: Array[String] = line.getAs("desc").toString.split("\\)")
      val descArray: Array[String] = str1(0).split(",")
      Topic_desc_result(line.getAs("id"), descArray)
  }
  //转化成word2Vec
  private val dframe: DataFrame = dframerdd.toDF()
  private val word2Vec: Word2Vec