Offline scheduling

Step 1: Prepare the dataset

create table if not exists ad (id STRING,age DOUBLE,sex DOUBLE,duration DOUBLE,place DOUBLE,ctr DOUBLE ) partitioned by (dt STRING) ;
alter table ad add if not exists partition (dt='20160919') partition (dt='20160920');

Step 2: Create a pipeline

1. Create and open a custom pipeline. For more information, see Create a custom pipeline.

2. Build the pipeline.

   1. In the component list on the left, under Source/Target, drag two Read Table components onto the canvas, and rename them to ad-1 and ad-2 respectively.

   2. From the component list on the left, navigate to the Data Preprocessing folder and drag two Normalization components to the canvas.

   3. From the component list on the left, navigate to Machine Learning > Binary Classification and drag the Binary Logistic Regression component to the canvas.

   4. From the component list on the left, navigate to the Machine Learning folder and drag the Predicted component to the canvas.

   5. In the component list on the left, drag the Write Table component under Source/Target onto the canvas, and rename it to ad_result-1.

   6. Connect the components to build a pipeline as shown in the following figure.

      

      No.	Description
      ①	Import data sources.
      ②	Preprocess data.
      ③	Train the model.
      ④	Make predictions.

3. Configure the component parameters.

   1. Click the ad-2 (training data source) and ad-1 (prediction data source) components on the canvas. In the right-side pane, configure the data sources for the pipeline.

      Tab	Parameter	Description
      Select Table	Table Name	Enter ad.
      	Partition	Select the Partition check box.
      	Parameter	Set this parameter to dt=@@{yyyyMMdd} to ensure that daily incremental data is used for prediction.
      Field Details	Source Table Fields	After you configure parameters on the Select Table tab, the system automatically populates the Source Table Fields. You do not need to configure this parameter.

   2. Click the Normalization-1 and Normalization-2 components on the canvas. On the Field Settings tab in the right-side pane, click Select Fields and select the DOUBLE and INT fields.

   3. Click the Binary Logistic Regression component on the canvas. In the right-side pane, configure its parameters. Configure only the following parameters and leave the others at their default values.

      Tab	Parameter	Description
      Field Settings	Feature Columns For Training	Select the age, sex, duration, and place columns.
      	Target Column	Select the ctr column.

   4. Click the Predicted component on the canvas. In the right-side pane, configure its parameters. Configure only the following parameters and leave the others at their default values.

      Tab	Parameter	Description
      Field Settings	Feature Columns	Select the age, sex, duration, and place columns.
      	Reserved Columns	Select the ctr column.

   5. Click the ad_result-1 component on the canvas. On the Select Table tab in the right-side pane, set Output Table Name to ad_result.

4. Click the Run icon in the upper-left corner of the canvas to run the pipeline.

5. After the pipeline finishes running, right-click the ad_result-1 component and choose View Data > Output to view the prediction results.

   The output columns are:

   • prediction_result: indicates whether an ad was clicked. A value of 1 means clicked, and 0 means not clicked.

   • prediction_score: indicates the probability that the ad is clicked.

Step 3: Perform offline scheduling

1. Use DataWorks to create, configure, and commit a PAI task. For more information, see Create and run a PAI Studio node.

   When you configure the scheduling task, set Time to run the prediction pipeline at midnight (00:00) every day. For more information, see Configure time properties.

2. On the task submission page, click O&M in the upper-right corner to view task logs. For more information, see Manage auto-triggered tasks.