Java 类名：com.alibaba.alink.operator.batch.clustering.GmmTrainBatchOp
Python 类名：GmmTrainBatchOp

功能介绍

混合模型（Mixture Model）是一个可以用来表示在总体分布中含有K个子分布的概率模型。换句话说，混合模型表示了观测数据在总体中的概率分布，它是一个由K个子分布组成的混合分布。
而高斯混合模型（Gaussian Mixture Model, GMM）可以用来表示在总体分布中含有K个高斯子分布的概率模型。它通常可以被用作分类模型。

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

| vectorCol | 向量列名 | 向量列对应的列名 | String | ✓ | 所选列类型为 [DENSE_VECTOR, SPARSE_VECTOR, STRING, VECTOR] | |

| epsilon | 收敛阈值 | 当两轮迭代的中心点距离小于epsilon时，算法收敛。 | Double | | | 1.0E-4 |

| k | 聚类中心点数量 | 聚类中心点数量 | Integer | | | 2 |

| maxIter | 最大迭代步数 | 最大迭代步数，默认为 100 | Integer | | [1, +inf) | 100 |

| randomSeed | 随机数种子 | 随机数种子 | Integer | | | 0 |

代码示例

Python 代码

  1. from pyalink.alink import *
  3. import pandas as pd
  5. useLocalEnv(1)
  7. df_data = pd.DataFrame([
  8.     ["-0.6264538 0.1836433"],
  9.     ["-0.8356286 1.5952808"],
  10.     ["0.3295078 -0.8204684"],
  11.     ["0.4874291 0.7383247"],
  12.     ["0.5757814 -0.3053884"],
  13.     ["1.5117812 0.3898432"],
  14.     ["-0.6212406 -2.2146999"],
  15.     ["11.1249309 9.9550664"],
  16.     ["9.9838097 10.9438362"],
  17.     ["10.8212212 10.5939013"],
  18.     ["10.9189774 10.7821363"],
  19.     ["10.0745650 8.0106483"],
  20.     ["10.6198257 9.9438713"],
  21.     ["9.8442045 8.5292476"],
  22.     ["9.5218499 10.4179416"],
  23. ])
  25. data = BatchOperator.fromDataframe(df_data, schemaStr='features string')
  26. dataStream = StreamOperator.fromDataframe(df_data, schemaStr='features string')
  28. gmm = GmmTrainBatchOp() \
  29.     .setVectorCol("features") \
  30.     .setEpsilon(0.)
  32. model = gmm.linkFrom(data)
  34. predictor = GmmPredictBatchOp() \
  35.     .setPredictionCol("cluster_id") \
  36.     .setVectorCol("features") \
  37.     .setPredictionDetailCol("cluster_detail")
  39. predictor.linkFrom(model, data).print()
  41. predictorStream = GmmPredictStreamOp(model) \
  42.     .setPredictionCol("cluster_id") \
  43.     .setVectorCol("features") \
  44.     .setPredictionDetailCol("cluster_detail")
  46. predictorStream.linkFrom(dataStream).print()
  48. StreamOperator.execute()

Java 代码

  1. import org.apache.flink.types.Row;
  3. import com.alibaba.alink.operator.batch.BatchOperator;
  4. import com.alibaba.alink.operator.batch.clustering.GmmPredictBatchOp;
  5. import com.alibaba.alink.operator.batch.clustering.GmmTrainBatchOp;
  6. import com.alibaba.alink.operator.batch.source.MemSourceBatchOp;
  7. import com.alibaba.alink.operator.stream.StreamOperator;
  8. import com.alibaba.alink.operator.stream.clustering.GmmPredictStreamOp;
  9. import com.alibaba.alink.operator.stream.source.MemSourceStreamOp;
  10. import org.junit.Test;
  12. import java.util.Arrays;
  13. import java.util.List;
  15. public class GmmTrainBatchOpTest {
  16.     @Test
  17.     public void testGmmTrainBatchOp() throws Exception {
  18.         List <Row> df_data = Arrays.asList(
  19.             Row.of("-0.6264538 0.1836433"),
  20.             Row.of("-0.8356286 1.5952808"),
  21.             Row.of("0.3295078 -0.8204684"),
  22.             Row.of("0.4874291 0.7383247"),
  23.             Row.of("0.5757814 -0.3053884"),
  24.             Row.of("1.5117812 0.3898432"),
  25.             Row.of("-0.6212406 -2.2146999"),
  26.             Row.of("11.1249309 9.9550664"),
  27.             Row.of("9.9838097 10.9438362"),
  28.             Row.of("10.8212212 10.5939013"),
  29.             Row.of("10.9189774 10.7821363"),
  30.             Row.of("10.0745650 8.0106483"),
  31.             Row.of("10.6198257 9.9438713"),
  32.             Row.of("9.8442045 8.5292476"),
  33.             Row.of("9.5218499 10.4179416")
  34.         );
  35.         BatchOperator <?> data = new MemSourceBatchOp(df_data, "features string");
  36.         StreamOperator <?> dataStream = new MemSourceStreamOp(df_data, "features string");
  37.         BatchOperator <?> gmm = new GmmTrainBatchOp()
  38.             .setVectorCol("features")
  39.             .setEpsilon(0.);
  40.         BatchOperator <?> model = gmm.linkFrom(data);
  41.         BatchOperator <?> predictor = new GmmPredictBatchOp()
  42.             .setPredictionCol("cluster_id")
  43.             .setVectorCol("features")
  44.             .setPredictionDetailCol("cluster_detail");
  45.         predictor.linkFrom(model, data).print();
  46.         StreamOperator <?> predictorStream = new GmmPredictStreamOp(model)
  47.             .setPredictionCol("cluster_id")
  48.             .setVectorCol("features")
  49.             .setPredictionDetailCol("cluster_detail");
  50.         predictorStream.linkFrom(dataStream).print();
  51.         StreamOperator.execute();
  52.     }
  53. }

运行结果

| features | cluster_id | cluster_detail | | —- | —- | —- |

| -0.6264538 0.1836433 | 1 | 4.275273913968281E-92 1.0 |

| -0.8356286 1.5952808 | 1 | 1.0260377730239899E-92 1.0 |

| 0.3295078 -0.8204684 | 1 | 1.0970173367545207E-80 1.0 |

| 0.4874291 0.7383247 | 1 | 3.302173132311E-75 1.0 |

| 0.5757814 -0.3053884 | 1 | 3.1638113605165424E-76 1.0 |

| 1.5117812 0.3898432 | 1 | 2.101805230873173E-62 1.0 |

| -0.6212406 -2.2146999 | 1 | 6.772270268600749E-97 1.0 |

| 11.1249309 9.9550664 | 0 | 1.0 3.156783801247968E-56 |

| 9.9838097 10.9438362 | 0 | 1.0 1.9024447346702425E-51 |

| 10.8212212 10.5939013 | 0 | 1.0 2.800973098729604E-56 |

| 10.9189774 10.7821363 | 0 | 1.0 1.7209132744891298E-57 |

| 10.0745650 8.0106483 | 0 | 1.0 2.8642696635130495E-43 |

| 10.6198257 9.9438713 | 0 | 1.0 5.773273991940433E-53 |

| 9.8442045 8.5292476 | 0 | 1.0 2.5273123050925483E-43 |

| 9.5218499 10.4179416 | 0 | 1.0 1.7314580596767853E-46 |