Platform For AI:バイナリ分類のロジスティック回帰

コンポーネントの設定

次のいずれかの方法を使用して、バイナリ分類のロジスティック回帰コンポーネントを設定できます。

方法1: Machine Learning Designerでコンポーネントを構成する

方法2: PAIコマンドによるコンポーネントの設定

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"

例：

1. 次のSQL文を実行してトレーニングデータを生成します。

   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
       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
       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
       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
       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
       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
   ) a;

   次のトレーニングデータテーブルlr_test_inputが生成されます。

   f0	f1	f2	f3	ラベル
   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. 次のPAIコマンドを実行して、バイナリ分類のロジスティック回帰コンポーネント用に設定されたトレーニングパラメータを送信します。

   lr_test_modelが存在する場合は、

   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. 次のPAIコマンドを実行して、予測コンポーネントに設定されたパラメーターを送信します。パラメーターの詳細については、「予測」をご参照ください。

   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. 予測結果テーブルlr_test_prediction_resultが生成されます。

   ラベル	prediction_result	prediction_score	予測_詳細
   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}