KerasSequential分类预测 (KerasSequentialClassifierPredictStreamOp)

Java 类名：com.alibaba.alink.operator.stream.classification.KerasSequentialClassifierPredictStreamOp

Python 类名：KerasSequentialClassifierPredictStreamOp

功能介绍

与 KerasSequential分类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
predictionDetailCol	预测详细信息列名	预测详细信息列名	String
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

source = CsvSourceBatchOp() \
    .setFilePath("https://alink-release.oss-cn-beijing.aliyuncs.com/data-files/random_tensor.csv") \
    .setSchemaStr("tensor string, label int")

source = ToTensorBatchOp() \
    .setSelectedCol("tensor") \
    .setTensorDataType("DOUBLE") \
    .setTensorShape([200, 3]) \
    .linkFrom(source)

streamSource = CsvSourceStreamOp() \
    .setFilePath("https://alink-release.oss-cn-beijing.aliyuncs.com/data-files/random_tensor.csv") \
    .setSchemaStr("tensor string, label int")

trainBatchOp = KerasSequentialClassifierTrainBatchOp() \
    .setTensorCol("tensor") \
    .setLabelCol("label") \
    .setLayers([
        "Conv1D(256, 5, padding='same', activation='relu')",
        "Conv1D(128, 5, padding='same', activation='relu')",
        "Dropout(0.1)",
        "MaxPooling1D(pool_size=8)",
        "Conv1D(128, 5, padding='same', activation='relu')",
        "Conv1D(128, 5, padding='same', activation='relu')",
        "Flatten()"
    ]) \
    .setOptimizer("Adam()") \
    .setNumEpochs(1) \
    .linkFrom(source)

predictStreamOp = KerasSequentialClassifierPredictStreamOp(trainBatchOp) \
    .setPredictionCol("pred") \
    .setPredictionDetailCol("pred_detail") \
    .setReservedCols(["label"]) \
    .linkFrom(streamSource)
predictStreamOp.print()
StreamOperator.execute()

Java 代码

import com.alibaba.alink.operator.batch.BatchOperator;
import com.alibaba.alink.operator.batch.classification.KerasSequentialClassifierTrainBatchOp;
import com.alibaba.alink.operator.batch.dataproc.ToTensorBatchOp;
import com.alibaba.alink.operator.batch.source.CsvSourceBatchOp;
import com.alibaba.alink.operator.stream.StreamOperator;
import com.alibaba.alink.operator.stream.classification.KerasSequentialClassifierPredictStreamOp;
import com.alibaba.alink.operator.stream.source.CsvSourceStreamOp;
import org.junit.Test;

public class KerasSequentialClassifierPredictStreamOpTest {

	@Test
	public void testKerasSequentialClassifierPredictStreamOp() throws Exception {
		BatchOperator <?> source = new CsvSourceBatchOp()
			.setFilePath("https://alink-release.oss-cn-beijing.aliyuncs.com/data-files/random_tensor.csv")
			.setSchemaStr("tensor string, label int");

		source = new ToTensorBatchOp()
			.setSelectedCol("tensor")
			.setTensorDataType("DOUBLE")
			.setTensorShape(200, 3)
			.linkFrom(source);

		StreamOperator <?> streamSource = new CsvSourceStreamOp()
			.setFilePath("https://alink-release.oss-cn-beijing.aliyuncs.com/data-files/random_tensor.csv")
			.setSchemaStr("tensor string, label int");

		KerasSequentialClassifierTrainBatchOp trainBatchOp = new KerasSequentialClassifierTrainBatchOp()
			.setTensorCol("tensor")
			.setLabelCol("label")
			.setLayers(new String[] {
				"Conv1D(256, 5, padding='same', activation='relu')",
				"Conv1D(128, 5, padding='same', activation='relu')",
				"Dropout(0.1)",
				"MaxPooling1D(pool_size=8)",
				"Conv1D(128, 5, padding='same', activation='relu')",
				"Conv1D(128, 5, padding='same', activation='relu')",
				"Flatten()"
			})
			.setOptimizer("Adam()")
			.setNumEpochs(1)
			.linkFrom(source);

		KerasSequentialClassifierPredictStreamOp predictStreamOp = new KerasSequentialClassifierPredictStreamOp(trainBatchOp)
			.setPredictionCol("pred")
			.setPredictionDetailCol("pred_detail")
			.setReservedCols("label")
			.linkFrom(streamSource);
		predictStreamOp.print();
		StreamOperator.execute();
	}
}

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}