阿里云人工智能平台PAI逻辑回归二分类-云淘科技

组件配置

您可以使用以下任意一种方式，配置逻辑回归二分类组件参数。

方式一：可视化方式

在Designer（原PAI-Studio）工作流页面配置组件参数。

方式二：PAI命令方式

使用PAI命令方式，配置该组件参数。您可以使用SQL脚本组件进行PAI命令调用，详情请参见SQL脚本。

PAI -name logisticregression_binary
    -project algo_public
    -DmodelName="xlab_m_logistic_regression_6096"
    -DregularizedLevel="1"
    -DmaxIter="100"
    -DregularizedType="l1"
    -Depsilon="0.000001"
    -DlabelColName="y"
    -DfeatureColNames="pdays,emp_var_rate"
    -DgoodValue="1"
    -DinputTableName="bank_data"

PAI使用KV格式表示稀疏数据，如下表所示。其中itemDelimiter表示KV对之间的分隔符，kvDelimiter表示key和value之间的分隔符。

说明

KV格式使用索引（从0开始）表示key。如果使用字符表示key，则系统报错。

示例

1. 使用SQL语句，生成训练数据。

   drop table if exists lr_test_input;
   create table lr_test_input
   as
   select
       *
   from
   (
   select
           cast(1 as double) as f0,
           cast(0 as double) as f1,
           cast(0 as double) as f2,
           cast(0 as double) as f3,
           cast(0 as bigint) as label
       from dual
       union all
           select
               cast(0 as double) as f0,
               cast(1 as double) as f1,
               cast(0 as double) as f2,
               cast(0 as double) as f3,
               cast(0 as bigint) as label
       from dual
       union all
           select
               cast(0 as double) as f0,
               cast(0 as double) as f1,
               cast(1 as double) as f2,
               cast(0 as double) as f3,
               cast(1 as bigint) as label
       from dual
       union all
           select
               cast(0 as double) as f0,
               cast(0 as double) as f1,
               cast(0 as double) as f2,
               cast(1 as double) as f3,
               cast(1 as bigint) as label
       from dual
       union all
           select
               cast(1 as double) as f0,
               cast(0 as double) as f1,
               cast(0 as double) as f2,
               cast(0 as double) as f3,
               cast(0 as bigint) as label
       from dual
       union all
           select
               cast(0 as double) as f0,
               cast(1 as double) as f1,
               cast(0 as double) as f2,
               cast(0 as double) as f3,
               cast(0 as bigint) as label
       from dual
   ) a;

   生成的训练数据表lr_test_input如下。

   f0	f1	f2	f3	label
   1.0	0.0	0.0	0.0	0
   0.0	0.0	1.0	0.0	1
   0.0	0.0	0.0	1.0	1
   0.0	1.0	0.0	0.0	0
   1.0	0.0	0.0	0.0	0
   0.0	1.0	0.0	0.0	0

2. 使用PAI命令，提交逻辑回归二分类组件的训练参数。

   drop offlinemodel if exists lr_test_model;
   PAI -name logisticregression_binary
       -project algo_public
       -DmodelName="lr_test_model"
       -DitemDelimiter=","
       -DregularizedLevel="1"
       -DmaxIter="100"
       -DregularizedType="None"
       -Depsilon="0.000001"
       -DkvDelimiter=":"
       -DlabelColName="label"
       -DfeatureColNames="f0,f1,f2,f3"
       -DenableSparse="false"
       -DgoodValue="1"
       -DinputTableName="lr_test_input";

3. 使用PAI命令，提交预测组件参数。更多参数说明，请参见预测。

   drop table if exists lr_test_prediction_result;
   PAI -name prediction
       -project algo_public
       -DdetailColName="prediction_detail"
       -DmodelName="lr_test_model"
       -DitemDelimiter=","
       -DresultColName="prediction_result"
       -Dlifecycle="28"
       -DoutputTableName="lr_test_prediction_result"
       -DscoreColName="prediction_score"
       -DkvDelimiter=":"
       -DinputTableName="lr_test_input"
       -DenableSparse="false"
       -DappendColNames="label";

4. 查看预测结果表lr_test_prediction_result。

   label	prediction_result	prediction_score	prediction_detail
   0	0	0.9999998793434426	{“0”: 0.9999998793434426, “1”: 1.206565574533681e-07}
   1	1	0.999999799574135	{“0”: 2.004258650156743e-07, “1”: 0.999999799574135}
   1	1	0.999999799574135	{“0”: 2.004258650156743e-07, “1”: 0.999999799574135}
   0	0	0.9999998793434426	{“0”: 0.9999998793434426, “1”: 1.206565574533681e-07}
   0	0	0.9999998793434426	{“0”: 0.9999998793434426, “1”: 1.206565574533681e-07}
   0	0	0.9999998793434426	{“0”: 0.9999998793434426, “1”: 1.206565574533681e-07}