83 sample

src/main/scala/com/aliyun/odps/spark/examples/makedata_recsys_r_rate/makedata_recsys_fm_sample_20250317.scala

@@ -0,0 +1,116 @@
+package com.aliyun.odps.spark.examples.makedata_recsys_r_rate
+
+import com.aliyun.odps.spark.examples.myUtils.{DataUtils, MyDateUtils, MyHdfsUtils, ParamUtils}
+import org.apache.hadoop.io.compress.GzipCodec
+import org.apache.spark.sql.SparkSession
+
+import scala.collection.JavaConversions._
+import scala.collection.mutable.ArrayBuffer
+import scala.util.Random
+
+object makedata_recsys_fm_sample_20250317 {
+  def main(args: Array[String]): Unit = {
+    // 1. 读取参数
+    val param = ParamUtils.parseArgs(args)
+    val readPath = param.getOrElse("readPath", "/dw/recommend/model/83_origin_data/")
+    val beginStr = param.getOrElse("beginStr", "20250317")
+    val endStr = param.getOrElse("endStr", "20250317")
+    val whatApps = param.getOrElse("whatApps", "0,4,5,21,3,6").split(",").toSet
+    val whatLabel = param.getOrElse("whatLabel", "is_return_n_noself")
+    val fuSampleRate = param.getOrElse("fuSampleRate", "-1.0").toDouble
+    val notUseBucket = param.getOrElse("notUseBucket", "0").toInt
+    val featureNameFile = param.getOrElse("featureName", "20241209_recsys_nor_name.txt")
+    val featureBucketFile = param.getOrElse("featureBucket", "20241209_recsys_nor_bucket.txt")
+    val repartition = param.getOrElse("repartition", "100").toInt
+    val savePath = param.getOrElse("savePath", "/dw/recommend/model/83_recsys_rov_train_data/")
+
+    val spark = SparkSession
+      .builder()
+      .appName(this.getClass.getName)
+      .getOrCreate()
+    val sc = spark.sparkContext
+
+    // 2. 加载特征
+    val loader = getClass.getClassLoader
+    val featureNameSet = DataUtils.loadUseFeatureNames(loader, featureNameFile)
+    val featureBucketMap = DataUtils.loadUseFeatureBuckets(loader, notUseBucket, featureBucketFile)
+    val bucketsMap_br = sc.broadcast(featureBucketMap)
+
+    // 3. 处理数据
+    val dateRange = MyDateUtils.getDateRange(beginStr, endStr)
+    for (date <- dateRange) {
+      println("开始执行:" + date)
+      val data = sc.textFile(readPath + "/" + date + "*").map(r => {
+          val rList = r.split("\t")
+          val logKey = rList(0)
+          val labelKey = rList(1)
+          val scoresMap = rList(2)
+          val featData = rList(3)
+          (logKey, labelKey, scoresMap, featData)
+        })
+        .filter {
+          case (logKey, labelKey, scoresMap, featData) =>
+            validData(logKey, whatApps)
+        }.filter {
+          case (logKey, labelKey, scoresMap, featData) =>
+            val label = DataUtils.parseLabel(labelKey, whatLabel).toDouble
+            label > 0 || new Random().nextDouble() <= fuSampleRate
+        }
+        .map {
+          case (logKey, labelKey, scoresMap, featData) =>
+            val label = DataUtils.parseLabel(labelKey, whatLabel).toDouble
+            val features = DataUtils.parseFeature(featData)
+            (logKey, label, scoresMap, features)
+        }
+        .mapPartitions(row => {
+          val result = new ArrayBuffer[String]()
+          val bucketsMap = bucketsMap_br.value
+          row.foreach {
+            case (logKey, label, scoresMap, features) =>
+              val featuresBucket = DataUtils.bucketFeature(featureNameSet, bucketsMap, features)
+              result.add(label + "\t" + featuresBucket.mkString("\t"))
+          }
+          result.iterator
+        })
+
+      // 4. 保存数据到hdfs
+      val hdfsPath = savePath + "/" + date
+      if (hdfsPath.nonEmpty && hdfsPath.startsWith("/dw/recommend/model/")) {
+        println("删除路径并开始数据写入:" + hdfsPath)
+        MyHdfsUtils.delete_hdfs_path(hdfsPath)
+        data.repartition(repartition).saveAsTextFile(hdfsPath, classOf[GzipCodec])
+      } else {
+        println("路径不合法，无法写入:" + hdfsPath)
+      }
+    }
+  }
+
+  private def recommendFlow(flowPool: String): Boolean = {
+    if (flowPool.isEmpty || flowPool.endsWith("#1")) {
+      return true
+    }
+    false
+  }
+
+  private def validData(logKey: String, whatApps: Set[String]): Boolean = {
+    // apptype, page, pagesource, recommendpagetype, flowpool, abcode, mid, vid, level, ts
+    val cells = logKey.split(",")
+    val apptype = cells(0)
+    val page = cells(1)
+    //val pagesource = cells(2)
+    val recommendpagetype = cells(3)
+    val flowpool = cells(4)
+    if (whatApps.contains(apptype)) {
+      if (recommendFlow(flowpool)) {
+        if (page.equals("详情后沉浸页")) {
+          return true
+        } else if (page.equals("回流后沉浸页&内页feed")) {
+          return true
+        } else if (page.equals("首页feed")) {
+          return true
+        }
+      }
+    }
+    false
+  }
+}

src/main/scala/com/aliyun/odps/spark/examples/myUtils/DataUtils.scala

@@ -3,11 +3,13 @@ package com.aliyun.odps.spark.examples.myUtils
 import com.alibaba.fastjson.{JSON, JSONObject}
 import com.aliyun.odps.TableSchema
 import com.aliyun.odps.data.Record
+import examples.extractor.ExtractorUtils
 import org.apache.hadoop.io.compress.GzipCodec
 import org.apache.spark.SparkContext
 import org.apache.spark.rdd.RDD
 
 import scala.collection.JavaConversions.mapAsScalaMap
+import scala.io.Source
 import scala.util.Random
 
 object DataUtils {
@@ -101,4 +103,77 @@ object DataUtils {
     }
     false
   }
+
+  def loadFileData(loader: ClassLoader, nameFile: String): String = {
+    val resourceUrlBucket = loader.getResource(nameFile)
+    val data =
+      if (resourceUrlBucket != null) {
+        val buckets = Source.fromURL(resourceUrlBucket).getLines().mkString("\n")
+        Source.fromURL(resourceUrlBucket).close()
+        buckets
+      } else {
+        ""
+      }
+    data
+  }
+
+  def loadUseFeatureNames(loader: ClassLoader, nameFile: String): Set[String] = {
+    val names = loadFileData(loader, nameFile)
+    println(names)
+    names.split("\n")
+      .map(r => r.replace(" ", "").replaceAll("\n", ""))
+      .filter(r => r.nonEmpty)
+      .toSet
+  }
+
+  def loadUseFeatureBuckets(loader: ClassLoader, notUseBucket: Int, nameFile: String): Map[String, (Double, Array[Double])] = {
+    if (notUseBucket > 0) {
+      return Map[String, (Double, Array[Double])]()
+    }
+    val buckets = loadFileData(loader, nameFile)
+    println(buckets)
+    buckets.split("\n")
+      .map(r => r.replace(" ", "").replaceAll("\n", ""))
+      .filter(r => r.nonEmpty)
+      .map(r => {
+        val rList = r.split("\t")
+        (rList(0), (rList(1).toDouble, rList(2).split(",").map(_.toDouble)))
+      }).toMap
+  }
+
+  def parseLabel(data: String, key: String, default: String = "0"): String = {
+    JSON.parseObject(data).getOrDefault(key, default).toString
+  }
+
+  def parseFeature(data: String): scala.collection.mutable.Map[String, Double] = {
+    val features = scala.collection.mutable.Map[String, Double]()
+    if (data.nonEmpty) {
+      val obj = JSON.parseObject(data)
+      obj.foreach(r => {
+        features.put(r._1, obj.getDoubleValue(r._1))
+      })
+    }
+    features
+  }
+
+  def bucketFeature(nameSet: Set[String], bucketMap: Map[String, (Double, Array[Double])], features: scala.collection.mutable.Map[String, Double]): Iterable[String] = {
+    features.map {
+      case (name, score) =>
+        if (!nameSet.contains(name)) {
+          ""
+        } else {
+          if (score > 1E-8) {
+            if (bucketMap.contains(name)) {
+              val (bucketsNum, buckets) = bucketMap(name)
+              val scoreNew = 1.0 / bucketsNum * (ExtractorUtils.findInsertPosition(buckets, score).toDouble + 1.0)
+              name + ":" + scoreNew.toString
+            } else {
+              name + ":" + score.toString
+            }
+          } else {
+            ""
+          }
+        }
+    }.filter(_.nonEmpty)
+  }
 }