Java 类名：com.alibaba.alink.operator.stream.regression.KerasSequentialRegressorPredictStreamOp
Python 类名：KerasSequentialRegressorPredictStreamOp

功能介绍

与 KerasSequential 回归训练组件对应的流预测组件。

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 | | —- | —- | —- | —- | —- | —- | —- |

| predictionCol | 预测结果列名 | 预测结果列名 | String | ✓ | | |

| inferBatchSize | 推理数据批大小 | 推理数据批大小 | Integer | | | 256 |

| modelFilePath | 模型的文件路径 | 模型的文件路径 | String | | | null |

| reservedCols | 算法保留列名 | 算法保留列 | String[] | | | null |

| modelStreamFilePath | 模型流的文件路径 | 模型流的文件路径 | String | | | null |

| modelStreamScanInterval | 扫描模型路径的时间间隔 | 描模型路径的时间间隔，单位秒 | Integer | | | 10 |

| modelStreamStartTime | 模型流的起始时间 | 模型流的起始时间。默认从当前时刻开始读。使用yyyy-mm-dd hh:mm:ss.fffffffff格式，详见Timestamp.valueOf(String s) | String | | | null |

代码示例

以下代码仅用于示意，可能需要修改部分代码或者配置环境后才能正常运行！

Python 代码

  1. source = CsvSourceBatchOp() \
  2.     .setFilePath("https://alink-release.oss-cn-beijing.aliyuncs.com/data-files/random_tensor.csv") \
  3.     .setSchemaStr("tensor string, label double")
  5. source = ToTensorBatchOp() \
  6.     .setSelectedCol("tensor") \
  7.     .setTensorDataType("DOUBLE") \
  8.     .setTensorShape([200, 3]) \
  9.     .linkFrom(source)
  11. streamSource = CsvSourceStreamOp() \
  12.     .setFilePath("https://alink-release.oss-cn-beijing.aliyuncs.com/data-files/random_tensor.csv") \
  13.     .setSchemaStr("tensor string, label double")
  15. trainBatchOp = KerasSequentialRegressorTrainBatchOp() \
  16.     .setTensorCol("tensor") \
  17.     .setLabelCol("label") \
  18.     .setLayers([
  19.         "Conv1D(256, 5, padding='same', activation='relu')",
  20.         "Conv1D(128, 5, padding='same', activation='relu')",
  21.         "Dropout(0.1)",
  22.         "MaxPooling1D(pool_size=8)",
  23.         "Conv1D(128, 5, padding='same', activation='relu')",
  24.         "Conv1D(128, 5, padding='same', activation='relu')",
  25.         "Flatten()"
  26.     ]) \
  27.     .setOptimizer("Adam()") \
  28.     .setNumEpochs(1) \
  29.     .linkFrom(source)
  31. predictStreamOp = KerasSequentialRegressorPredictStreamOp(trainBatchOp) \
  32.     .setPredictionCol("pred") \
  33.     .setReservedCols(["label"]) \
  34.     .linkFrom(streamSource)
  35. predictStreamOp.print()
  36. StreamOperator.execute()

Java 代码

  1. import com.alibaba.alink.operator.batch.BatchOperator;
  2. import com.alibaba.alink.operator.batch.dataproc.ToTensorBatchOp;
  3. import com.alibaba.alink.operator.batch.regression.KerasSequentialRegressorTrainBatchOp;
  4. import com.alibaba.alink.operator.batch.source.CsvSourceBatchOp;
  5. import com.alibaba.alink.operator.stream.StreamOperator;
  6. import com.alibaba.alink.operator.stream.regression.KerasSequentialRegressorPredictStreamOp;
  7. import com.alibaba.alink.operator.stream.source.CsvSourceStreamOp;
  8. import org.junit.Test;
  10. public class KerasSequentialRegressorPredictStreamOpTest {
  12.     @Test
  13.     public void testKerasSequentialRegressorPredictStreamOp() throws Exception {
  14.         BatchOperator <?> source = new CsvSourceBatchOp()
  15.             .setFilePath("https://alink-release.oss-cn-beijing.aliyuncs.com/data-files/random_tensor.csv")
  16.             .setSchemaStr("tensor string, label double");
  18.         source = new ToTensorBatchOp()
  19.             .setSelectedCol("tensor")
  20.             .setTensorDataType("DOUBLE")
  21.             .setTensorShape(200, 3)
  22.             .linkFrom(source);
  24.         StreamOperator <?> streamSource = new CsvSourceStreamOp()
  25.             .setFilePath("https://alink-release.oss-cn-beijing.aliyuncs.com/data-files/random_tensor.csv")
  26.             .setSchemaStr("tensor string, label double");
  28.         KerasSequentialRegressorTrainBatchOp trainBatchOp = new KerasSequentialRegressorTrainBatchOp()
  29.             .setTensorCol("tensor")
  30.             .setLabelCol("label")
  31.             .setLayers(new String[] {
  32.                 "Conv1D(256, 5, padding='same', activation='relu')",
  33.                 "Conv1D(128, 5, padding='same', activation='relu')",
  34.                 "Dropout(0.1)",
  35.                 "MaxPooling1D(pool_size=8)",
  36.                 "Conv1D(128, 5, padding='same', activation='relu')",
  37.                 "Conv1D(128, 5, padding='same', activation='relu')",
  38.                 "Flatten()"
  39.             })
  40.             .setOptimizer("Adam()")
  41.             .setNumEpochs(1)
  42.             .linkFrom(source);
  44.         KerasSequentialRegressorPredictStreamOp predictStreamOp = new KerasSequentialRegressorPredictStreamOp(trainBatchOp)
  45.             .setPredictionCol("pred")
  46.             .setReservedCols("label")
  47.             .linkFrom(streamSource);
  48.         predictStreamOp.print();
  49.         StreamOperator.execute();
  50.     }
  51. }

运行结果

| label | pred | pred_detail | | —- | —- | —- |

| 0 | 0 | {“0”:0.636155836712713,”1”:0.36384416328728697} |

| 1 | 0 | {“0”:0.6334926095655181,”1”:0.3665073904344819} |

| 1 | 0 | {“0”:0.6381823204965642,”1”:0.3618176795034358} |

| 1 | 0 | {“0”:0.6376416296248051,”1”:0.362358370375195} |

| 1 | 0 | {“0”:0.6345856985385896,”1”:0.36541430146141035} |

| 1 | 0 | {“0”:0.6357593109428179,”1”:0.364240689057182} |

| 0 | 0 | {“0”:0.6404387449594703,”1”:0.3595612550405296} |

| 1 | 0 | {“0”:0.6372702905549685,”1”:0.36272970944503136} |

| 0 | 0 | {“0”:0.635502012172225,”1”:0.36449798782777487} |

| 0 | 0 | {“0”:0.6262401788033837,”1”:0.37375982119661644} |