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

功能介绍

二分k均值算法是k-means聚类算法的一个变体，主要是为了改进k-means算法随机选择初始质心的随机性造成聚类结果不确定性的问题。
二分k均值方法是一种自顶向下的聚类。初始阶段该算法将所有节点视为属于同一个cluster。在每次迭代中，它选择一个类别，然后使用Kmeans算法对该类中的所有数据点进行二分类。当没有类可拆分时或者达到最大聚类个数时，算法终止。

参数说明

训练

| 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] | |

| distanceType | 距离度量方式 | 聚类使用的距离类型 | String | | “EUCLIDEAN”, “COSINE” | “EUCLIDEAN” |

| k | 聚类中心点数目 | 聚类中心点数目 | Integer | | | 4 |

| maxIter | 最大迭代步数 | 最大迭代步数，默认为 10。 | Integer | | | 10 |

| minDivisibleClusterSize | 最小可分裂的聚类大小 | 最小可分裂的聚类大小 | Integer | | | 1 |

| 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 = pd.DataFrame([
  8.     [0, "0 0 0"],
  9.     [1, "0.1,0.1,0.1"],
  10.     [2, "0.2,0.2,0.2"],
  11.     [3, "9 9 9"],
  12.     [4, "9.1 9.1 9.1"],
  13.     [5, "9.2 9.2 9.2"]
  14. ])
  16. inBatch = BatchOperator.fromDataframe(df, schemaStr='id int, vec string')
  17. inStream = StreamOperator.fromDataframe(df, schemaStr='id int, vec string')
  19. kmeansTrain = BisectingKMeansTrainBatchOp()\
  20.     .setVectorCol("vec")\
  21.     .setK(2)\
  22.     .linkFrom(inBatch)
  23. kmeansTrain.lazyPrint(10)
  25. predictBatch = BisectingKMeansPredictBatchOp()\
  26.     .setPredictionCol("pred")\
  27.     .linkFrom(kmeansTrain, inBatch)
  28. predictBatch.print()
  30. predictStream = BisectingKMeansPredictStreamOp(kmeansTrain)\
  31.     .setPredictionCol("pred")\
  32.     .linkFrom(inStream)
  33. predictStream.print()
  35. 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.BisectingKMeansPredictBatchOp;
  5. import com.alibaba.alink.operator.batch.clustering.BisectingKMeansTrainBatchOp;
  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.BisectingKMeansPredictStreamOp;
  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 BisectingKMeansTrainBatchOpTest {
  16.     @Test
  17.     public void testBisectingKMeansTrainBatchOp() throws Exception {
  18.         List <Row> df = Arrays.asList(
  19.             Row.of(0, "0 0 0"),
  20.             Row.of(1, "0.1,0.1,0.1"),
  21.             Row.of(2, "0.2,0.2,0.2"),
  22.             Row.of(3, "9 9 9"),
  23.             Row.of(4, "9.1 9.1 9.1"),
  24.             Row.of(5, "9.2 9.2 9.2")
  25.         );
  26.         BatchOperator <?> inBatch = new MemSourceBatchOp(df, "id int, vec string");
  27.         StreamOperator <?> inStream = new MemSourceStreamOp(df, "id int, vec string");
  28.         BatchOperator <?> kmeansTrain = new BisectingKMeansTrainBatchOp()
  29.             .setVectorCol("vec")
  30.             .setK(2)
  31.             .linkFrom(inBatch);
  32.         kmeansTrain.lazyPrint(10);
  34.         BatchOperator <?> predictBatch = new BisectingKMeansPredictBatchOp()
  35.             .setPredictionCol("pred")
  36.             .linkFrom(kmeansTrain, inBatch);
  37.         predictBatch.print();
  39.         StreamOperator <?> predictStream = new BisectingKMeansPredictStreamOp(kmeansTrain)
  40.             .setPredictionCol("pred")
  41.             .linkFrom(inStream);
  42.         predictStream.print();
  43.         StreamOperator.execute();
  44.     }
  45. }

运行结果

模型结果

| model_id | model_info | | —- | —- |

| 0 | {“vectorCol”:””vec””,”distanceType”:””EUCLIDEAN””,”k”:”2”,”vectorSize”:”3”} |

| 1048576 | {“clusterId”:1,”size”:6,”center”:{“data”:[4.6,4.6,4.6]},”cost”:364.61999999999995} |

| 2097152 | {“clusterId”:2,”size”:3,”center”:{“data”:[0.1,0.1,0.1]},”cost”:0.06} |

| 3145728 | {“clusterId”:3,”size”:3,”center”:{“data”:[9.099999999999998,9.099999999999998,9.099999999999998]},”cost”:0.060000000000172804} |

预测结果

| id | vec | pred | | —- | —- | —- |

| 0 | 0 0 0 | 0 |

| 1 | 0.1,0.1,0.1 | 0 |

| 2 | 0.2,0.2,0.2 | 0 |

| 3 | 9 9 9 | 1 |

| 4 | 9.1 9.1 9.1 | 1 |

| 5 | 9.2 9.2 9.2 | 1 |