Manage training

Available in Classic and VPC

The Managing training section describes how to connect datasets to a service, run training, test completed training results, and change periodic training settings.

Connecting a dataset

To perform training, you must connect a dataset to the service.

To connect a dataset :

1. Navigate to > Services > AI Services > AiTEMS in the NAVER Cloud Platform console.
2. Click the Service menu.
3. In the service list, click the service to which you want to connect a dataset.
4. In Dataset information, click the [Manage dataset] button.
5. When the Manage dataset popup appears, select the dataset to connect from each dropdown menu.
6. Click [OK] button.
   • The connected dataset information appears in Dataset information, and the [Manage training] button becomes active.

Running training

To run dataset training:

1. Navigate to > Services > AI Services > AiTEMS in the NAVER Cloud Platform console.
2. Click the Service menu.
3. In the service list, click the service for which you want to run training.
4. Click the [Manage training] button.
   • The [Manage training] button is active only when datasets are connected. If the button is inactive, connect the datasets first.
5. When the Manage training popup appears, configure the training settings and click the [Start training] button.

   • Recommendation type: Select the recommendation type.

     • Personalized recommendations: Provides recommended items per user based on historical behavior.
     • Related item recommendations: Provides recommended items based on the related item history.
     • Popular item recommendations: Provides recommended items based on ranking and item popularity.

   • HPO settings: Configure options related to model generation.

     • If you do not configure HPO (Hyperparameter Optimization), AiTEMS uses its internal optimized algorithm for training. Configure HPO to achieve an optimal training model.
     Note

     For the Popular item recommendation type, HPO is mandatory. The group_column and max_group_recommend_top options can only be configured when selecting the Popular item recommendation type. The HPO parameters available for each recommendation type are as follows.

     Recommendation type	Parameter name	Description	Default value	Configurable range
     Personalized recommendations & Related item recommendations	session_max_timestamp	Maximum idle time (in seconds) between actions to consider as a single session	1,800 (30 minutes)	5 or higher
     	item_top_n	Maximum number of items used for recommendations, sorted by popularity	20,000	1,000~500,000
     	min_item_cut	Minimum number of occurrences required for an item	5	1 or higher
     	min_session_length	Minimum number of actions required within a session	2	2 or higher
     	max_user_recommend_top	Maximum number of recommended items per user	100	1~1,000
     	max_relate_recommend_top	Maximum number of related recommended items per item	100	1~1,000
     	remove_history_item	Whether to exclude items the user has previously interacted with from personalized recommendations	False	-
     Popular item recommendations	group_column	List of columns used to group items for popularity extraction	-	-
     	target_column	Column used as the target of popularity extraction	-	-
     	max_group_recommend_top	Maximum number of popular items per group	100	1~1,000

   • Training description: Enter a description for the training.

6. In the notification popup, click [OK] button.
   • You can check the training progress in the Training information status.

Testing training results

You can test the results to verify that the dataset training was completed correctly.
To test training results:

1. Navigate to > Services > AI Services > AiTEMS in the NAVER Cloud Platform console.
2. Click the Service menu.
3. In the service list, click the service whose training results you want to test.
4. Click the [Test recommendations] button.
5. When the Test recommendations popup appears, select the training to test.
6. Enter the required information depending on the selected training type.
   • Target ID: Enter the following information when testing personalized or related item recommendations.
     • Personalized recommendations: Enter the USER_ID to look up in the training results.
     • Related item recommendations: Enter the ITEM_ID to look up in the training results.
   • Column: When testing popular item recommendations, enter the data to look up in the training results.
7. Click the [Search] button.
   • You can check the test results in the result section at the bottom of the popup.

Changing periodic training settings

You can modify the update interval for periodic training and change the training configuration status to stop or restart training.

Modifying the update interval

The default update interval is 12 hours, and you can change it to 1/3/6/12 hours.
To modify the update interval:

1. Navigate to > Services > AI Services > AiTEMS in the NAVER Cloud Platform console.
2. Click the Service menu.
3. In the service list, click the service whose update interval you want to modify.
4. In Update interval, click the [Edit] button.
5. Select the new update interval and click the [Edit] button.

Changing the training configuration status

If training is in progress, you can change the status to Not configured to stop training.

To change the training configuration status:

1. Navigate to > Services > AI Services > AiTEMS in the NAVER Cloud Platform console.
2. Click the Service menu.
3. n the service list, click the service for which you want to change the training configuration status.
4. In Training configuration status, click either or .
   • indicates Not configured. Clicking it changes the status to Configured.
   • indicates Configured. Clicking it changes the status to Not configured.
5. If you changed the status from Not configured to Configured, the Manage training popup appears. Configure the training settings and click the [Start training] button.
   • For details on configuring training information, see Running training.
   • The training configuration status changes to Configured, and periodic training restarts.
6. If you changed the status from Configured to Not configured, review the notification popup and click the [OK] button.
   • The training configuration status changes to Not configured, and periodic training stops.