分层随机采样 (StratifiedSampleBatchOp)

Java 类名：com.alibaba.alink.operator.batch.dataproc.StratifiedSampleBatchOp

Python 类名：StratifiedSampleBatchOp

功能介绍

分层采样组件。给定输入数据，本算法根据用户指定的不同类别的采样比例进行随机采样。

当输入数据的所有类别均已通过ratios指定时，算法通过ratios中的比例进行采样； 当输入数据中存在类别没有通过ratios指定时，算法将使用ratio参数来采样。

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
strataCol	分层列	分层列	String	✓
strataRatios	采用比率	采用比率, eg, a:0.1,b:0.3	String	✓
strataRatio	采用比率	采用比率	Double			-1.0
withReplacement	是否放回	是否有放回的采样，默认不放回	Boolean			false

代码示例

Python 代码

from pyalink.alink import *

import pandas as pd

useLocalEnv(1)

df = pd.DataFrame([
        ['a',0.0,0.0],
        ['a',0.2,0.1],
        ['b',0.2,0.8],
        ['b',9.5,9.7],
        ['b',9.1,9.6],
        ['b',9.3,9.9]
    ])


batchData = BatchOperator.fromDataframe(df, schemaStr='x1 string, x2 double, x3 double')
sampleOp = StratifiedSampleBatchOp()\
       .setStrataCol("x1")\
       .setStrataRatios("a:0.5,b:0.5")

batchData.link(sampleOp).print()

Java 代码

import org.apache.flink.types.Row;

import com.alibaba.alink.operator.batch.BatchOperator;
import com.alibaba.alink.operator.batch.dataproc.StratifiedSampleBatchOp;
import com.alibaba.alink.operator.batch.source.MemSourceBatchOp;
import org.junit.Test;

import java.util.Arrays;
import java.util.List;

public class StratifiedSampleBatchOpTest {
	@Test
	public void testStratifiedSampleBatchOp() throws Exception {
		List <Row> df = Arrays.asList(
			Row.of("a", 0.0, 0.0),
			Row.of("a", 0.2, 0.1),
			Row.of("b", 0.2, 0.8),
			Row.of("b", 9.5, 9.7),
			Row.of("b", 9.1, 9.6),
			Row.of("b", 9.3, 9.9)
		);
		BatchOperator <?> batchData = new MemSourceBatchOp(df, "x1 string, x2 double, x3 double");
		BatchOperator <?> sampleOp = new StratifiedSampleBatchOp()
			.setStrataCol("x1")
			.setStrataRatios("a:0.5,b:0.5");
		batchData.link(sampleOp).print();
	}
}

运行结果

x1	x2	x3
a	0.0000	0.0000
b	9.5000	9.7000
b	9.1000	9.6000
b	9.3000	9.9000