:Custom metric configuration

Configure metrics

Create a metric

To create a metric, you must specify parameters such as a registration data table for which data registration and field configuration have been completed, a metric name, and a recommendation scenario.

1. Log in to the PAI-Rec console. In the left-side navigation pane, choose Metric management > Metric configuration.

2. Create a base metric.

   1. In the Base metric section, click Create and configure the following metrics: PV, UV, click, click_UV, ctr_sample_v1, and ctr_user_v1.

      • PV: Page views (PV)

      • UV: Unique visitors (UV)

        

      • click: Clicks

      • click_UV: Unique visitors (UV) who click

        

      • ctr_user_v1: Significance calculated based on the user dimension

      • ctr_sample_v1: Significance calculated based on the sample dimension

   2. Click Save.

3. Create a derived metric.

   1. In the Derived metric section, click Create and configure the uv_click_ctr metric.

      Set Metric name to uv_click_ctr, enter User click-through rate in the Description field, and select offline for Metric timeliness. In the Metric definition section, select click_UV, the division operator (/), and UV in sequence.

   2. Click Save.

Modify a metric

In the metric list, you can filter created metrics by recommendation scenario and timeliness. To modify a metric, click Details on the right of the target metric, and then click Modify. Modify the parameters as needed and click Save.

Calculate a metric

In the metric list, click Calculate on the right of the target metric. Set the time window and specific metrics, and then click OK to create a calculation task for the metric.

Delete a metric

In the metric list, click Delete on the right of the target metric.

Configure metric groups

You can group multiple metrics within a scenario. This lets you perform tasks, such as metric calculation, on the entire metric group.

Scheduled metric calculation

Each day, PAI-Rec dynamically assembles SQL statements based on your configured metrics. It reads data from the experiment report source table to calculate the metrics and stores the results in a specified Hologres table. For real-time metrics, the calculation job runs every 5 minutes.