Java 类名：com.alibaba.alink.operator.batch.outlier.IForestModelOutlierTrainBatchOp
Python 类名：IForestModelOutlierTrainBatchOp

功能介绍

iForest 可以识别数据中异常点，在异常检测领域有比较好的效果。算法使用 sub-sampling 方法，降低了算法的计算复杂度。

文献或出处

  1. Isolation Forest

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

| featureCols | 特征列名数组 | 特征列名数组，默认全选 | String[] | | 所选列类型为 [BIGDECIMAL, BIGINTEGER, BYTE, DOUBLE, FLOAT, INTEGER, LONG, SHORT] | null |

| numTrees | 模型中树的棵数 | 模型中树的棵数 | Integer | | | 100 |

| subsamplingSize | 每棵树的样本采样行数 | 每棵树的样本采样行数，默认 256 ，最小 2 ，最大 100000 . | Integer | | [1, 100000] | 256 |

| tensorCol | tensor列 | tensor列 | String | | 所选列类型为 [BOOL_TENSOR, BYTE_TENSOR, DOUBLE_TENSOR, FLOAT_TENSOR, INT_TENSOR, LONG_TENSOR, STRING, STRING_TENSOR, TENSOR, UBYTE_TENSOR] | null |

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

代码示例

Python 代码

  1. import pandas as pd
  2. df = pd.DataFrame([
  3. [0.73, 0],
  4. [0.24, 0],
  5. [0.63, 0],
  6. [0.55, 0],
  7. [0.73, 0],
  8. [0.41, 0]
  9. ])
  11. dataOp = BatchOperator.fromDataframe(df, schemaStr='val double, label int')
  13. trainOp = IForestModelOutlierTrainBatchOp()\
  14. .setFeatureCols(["val"])
  16. predOp = IForestModelOutlierPredictBatchOp()\
  17. .setOutlierThreshold(3.0)\
  18. .setPredictionCol("pred")\
  19. .setPredictionDetailCol("pred_detail")
  21. predOp.linkFrom(trainOp.linkFrom(dataOp), dataOp)
  23. evalOp = EvalOutlierBatchOp()\
  24. .setLabelCol("label")\
  25. .setPredictionDetailCol("pred_detail")\
  26. .setOutlierValueStrings(["1"]);
  28. metrics = predOp\
  29. .link(evalOp)\
  30. .collectMetrics()
  32. print(metrics)

Java 代码

  1. package com.alibaba.alink.operator.batch.outlier;
  3. import com.alibaba.alink.operator.batch.BatchOperator;
  4. import com.alibaba.alink.operator.batch.evaluation.EvalOutlierBatchOp;
  5. import com.alibaba.alink.operator.batch.source.MemSourceBatchOp;
  6. import com.alibaba.alink.operator.common.evaluation.OutlierMetrics;
  7. import com.alibaba.alink.testutil.AlinkTestBase;
  8. import org.junit.Assert;
  9. import org.junit.Test;
  11. public class IForestModelOutlierTrainBatchOpTest extends AlinkTestBase {
  13.     @Test
  14.     public void test() {
  15.         BatchOperator <?> data = new MemSourceBatchOp(
  16.             new Object[][] {
  17.                 {0.73, 0},
  18.                 {0.24, 0},
  19.                 {0.63, 0},
  20.                 {0.55, 0},
  21.                 {0.73, 0},
  22.                 {0.41, 0},
  23.             },
  24.             new String[]{"val", "label"});
  26.         IForestModelOutlierTrainBatchOp trainOp = new IForestModelOutlierTrainBatchOp()
  27.             .setFeatureCols("val");
  29.         IForestModelOutlierPredictBatchOp predOp = new IForestModelOutlierPredictBatchOp()
  30.             .setOutlierThreshold(3.0)
  31.             .setPredictionCol("pred")
  32.             .setPredictionDetailCol("pred_detail");
  34.         predOp.linkFrom(trainOp.linkFrom(data), data);
  36.         EvalOutlierBatchOp eval = new EvalOutlierBatchOp()
  37.             .setLabelCol("label")
  38.             .setPredictionDetailCol("pred_detail")
  39.             .setOutlierValueStrings("1");
  41.         OutlierMetrics metrics = predOp
  42.             .link(eval)
  43.             .collectMetrics();
  45.         Assert.assertEquals(1.0, metrics.getAccuracy(), 10e-6);
  46.     }
  47. }

运行结果

———————————————— Metrics: ————————————————
Outlier values: [1] Normal values: [0]
Auc:NaN Accuracy:1 Precision:1 Recall:0 F1:0

| Pred\Real | Outlier | Normal | | —- | —- | —- |

| Outlier | 0 | 0 |

| Normal | 0 | 6 |