Java 类名：com.alibaba.alink.operator.stream.recommendation.RecommendationRankingStreamOp
Python 类名：RecommendationRankingStreamOp

功能介绍

该组件功能是对召回的结果进行排序，并输出排序后的TopK个object，此处排序算法用户可以通过创建PipelineModel的方式定制，具体使用方式参见代码示例。

参数说明

名称 中文名称 描述 类型 是否必须？ 取值范围 默认值
mTableCol Not available! Not available! String ✓ Selected column type is [M_TABLE]
modelFilePath 模型的文件路径 Model file path String null
outputCol 输出结果列 Output result column name, optional, default null. String null
rankingCol 用来排序的得分列 Score column used for sorting. String null
reservedCols 算法保留列名 Algorithm reserved columns String[] null
topN 前N的数据 Select the closest N data Integer [1, +inf) 10
modelStreamFilePath 模型流的文件路径 File path of the model stream String null
modelStreamScanInterval 扫描模型路径的时间间隔 Time interval for scanning the model path, in seconds. Integer 10
modelStreamStartTime 模型流的起始时间 Start time of the model stream. Default is to read from the current time. Use yyyy-mm-dd hh:mm:ss.fffffffff format, see Timestamp.valueOf(String s) for details. String null

代码示例

Python 代码

  1. from pyalink.alink import *
  3. import pandas as pd
  5. useLocalEnv(1)
  7. import pandas as pd
  9. data = pd.DataFrame([["u6", "0.0 1.0", 0.0, 1.0, 1, "{\"data\":{\"iid\":[18,19,88]},\"schema\":\"iid INT\"}"]])
  10. predData = StreamOperator.fromDataframe(data, schemaStr='uid string, uf string, f0 double, f1 double, labels int, ilist string')
  11. predData = predData.link(ToMTableStreamOp().setSelectedCol("ilist"))
  12. data = pd.DataFrame([
  13.             ["u0", "1.0 1.0", 1.0, 1.0, 1, 18],
  14.             ["u1", "1.0 1.0", 1.0, 1.0, 0, 19],
  15.             ["u2", "1.0 0.0", 1.0, 0.0, 1, 88],
  16.             ["u3", "1.0 0.0", 1.0, 0.0, 0, 18],
  17.             ["u4", "0.0 1.0", 0.0, 1.0, 1, 88],
  18.             ["u5", "0.0 1.0", 0.0, 1.0, 0, 19],
  19.             ["u6", "0.0 1.0", 0.0, 1.0, 1, 88]]);
  20. trainData = BatchOperator.fromDataframe(data, schemaStr='uid string, uf string, f0 double, f1 double, labels int, iid string')
  21. oneHotCols = ["uid", "f0", "f1", "iid"]
  22. multiHotCols = ["uf"]
  23. pipe = Pipeline() \
  24.     .add( \
  25.         OneHotEncoder() \
  26.             .setSelectedCols(oneHotCols) \
  27.             .setOutputCols(["ovec"])) \
  28.     .add( \
  29.         MultiHotEncoder().setDelimiter(" ") \
  30.             .setSelectedCols(multiHotCols) \
  31.             .setOutputCols(["mvec"])) \
  32.     .add( \
  33.         VectorAssembler() \
  34.             .setSelectedCols(["ovec", "mvec"]) \
  35.             .setOutputCol("vec")) \
  36.     .add(
  37.         LogisticRegression() \
  38.             .setVectorCol("vec") \
  39.             .setLabelCol("labels") \
  40.             .setReservedCols(["uid", "iid"]) \
  41.             .setPredictionDetailCol("detail") \
  42.             .setPredictionCol("pred")) \
  43.     .add( \
  44.         JsonValue() \
  45.             .setSelectedCol("detail") \
  46.             .setJsonPath(["$.1"]) \
  47.             .setOutputCols(["score"]))
  48. lrModel = pipe.fit(trainData)
  49. rank = RecommendationRankingStreamOp(lrModel.save())\
  50.             .setMTableCol("ilist")\
  51.             .setOutputCol("il")\
  52.             .setTopN(2)\
  53.             .setRankingCol("score")\
  54.             .setReservedCols(["uid", "labels"])
  55. rank.linkFrom(predData).print()
  56. StreamOperator.execute()

Java 代码

  1. import org.apache.flink.types.Row;
  3. import com.alibaba.alink.operator.batch.StreamOperator;
  4. import com.alibaba.alink.operator.batch.source.MemSourceStreamOp;
  5. import com.alibaba.alink.pipeline.Pipeline;
  6. import com.alibaba.alink.pipeline.classification.LogisticRegression;
  7. import com.alibaba.alink.pipeline.dataproc.JsonValue;
  8. import com.alibaba.alink.pipeline.dataproc.vector.VectorAssembler;
  9. import com.alibaba.alink.pipeline.feature.MultiHotEncoder;
  10. import com.alibaba.alink.pipeline.feature.OneHotEncoder;
  11. import org.junit.Test;
  13. import java.util.Arrays;
  15. public class RecommendationRankingTest {
  17.      @Test
  18.      public void test() throws Exception {
  20.         Row[] predArray = new Row[] {
  21.             Row.of("u6", "0.0 1.0", 0.0, 1.0, 1, "{\"data\":{\"iid\":[18,19,88]},"
  22.                 + "\"schema\":\"iid INT\"}")
  23.         };
  25.         Row[] trainArray = new Row[] {
  26.             Row.of("u0", "1.0 1.0", 1.0, 1.0, 1, 18),
  27.             Row.of("u1", "1.0 1.0", 1.0, 1.0, 0, 19),
  28.             Row.of("u2", "1.0 0.0", 1.0, 0.0, 1, 88),
  29.             Row.of("u3", "1.0 0.0", 1.0, 0.0, 1, 18),
  30.             Row.of("u4", "0.0 1.0", 0.0, 1.0, 1, 88),
  31.             Row.of("u5", "0.0 1.0", 0.0, 1.0, 1, 19),
  32.             Row.of("u6", "0.0 1.0", 0.0, 1.0, 1, 88)
  33.         };
  34.         BatchOperator <?> trainData = new MemSourceBatchOp(Arrays.asList(trainArray),
  35.                 new String[] {"uid", "uf", "f0", "f1", "labels", "iid"});
  36.         StreamOperator <?> predData =  new MemSourceStreamOp(Arrays.asList(predArray),
  37.                 new String[] {"uid", "uf", "f0", "f1", "labels", "ilist"});
  39.         String[] oneHotCols = new String[] {"uid", "f0", "f1", "iid"};
  40.         String[] multiHotCols = new String[] {"uf"};
  42.         Pipeline pipe = new Pipeline()
  43.             .add(
  44.                 new OneHotEncoder()
  45.                     .setSelectedCols(oneHotCols)
  46.                     .setOutputCols("ovec"))
  47.             .add(
  48.                 new MultiHotEncoder().setDelimiter(" ")
  49.                     .setSelectedCols(multiHotCols)
  50.                     .setOutputCols("mvec"))
  51.             .add(
  52.                 new VectorAssembler()
  53.                     .setSelectedCols("ovec", "mvec")
  54.                     .setOutputCol("vec"))
  55.             .add(
  56.                 new LogisticRegression()
  57.                     .setVectorCol("vec")
  58.                     .setLabelCol("labels")
  59.                     .setReservedCols("uid", "iid")
  60.                     .setPredictionDetailCol("detail")
  61.                     .setPredictionCol("pred"))
  62.             .add(
  63.                 new JsonValue()
  64.                     .setSelectedCol("detail")
  65.                     .setJsonPath("$.1")
  66.                     .setOutputCols("score"));
  67.         RecommendationRankingStreamOp rank = new RecommendationRankingStreamOp(pipe.fit(trainData).save())
  68.                 .setMTableCol("ilist")
  69.                 .setOutputCol("ilist")
  70.                 .setTopN(2)
  71.                 .setRankingCol("score")
  72.                 .setReservedCols("uid", "labels");
  73.         rank.linkFrom(predData).print();
  74.         StreamOperator.execute();
  75.     }
  76. }

运行结果

uid uf f0 f1 labels ilist
u6 0.0 1.0 0.0000 1.0000 1 {“data”:{“iid”:[18,88],”score”:[0.9999999999999553,0.9999999999999472]},”schema”:”iid INT,score DOUBLE”}