- Print
- PDF
Learning management
- Print
- PDF
Available in Classic and VPC
Learning management describes how to connect a dataset to be learned on to the service, how to proceed with learning, how to test completed learning results, and how to change cycle learning settings.
Connect dataset
To proceed with learning, you need to connect a dataset to the service.
- To connect a dataset to the service, at least 1 dataset created for each type must exist. For more details on how to create a dataset, see Create dataset.
- You can also change the dataset linked to the service in the same way as linking datasets. However, the dataset cannot be changed while the service status is scheduled for learning or learning. Change the dataset after learning is complete, or change the dataset after stopping learning.
The following describes how to connect a dataset.
- Click the environment you are using in the Region menu and Platform menu of the NAVER Cloud Platform console.
- Click the Services > AI Services > AiTEMS menus in that order.
- Click the Service menu.
- Click the service to connect the dataset to from the list of services.
- Click the [Dataset management] button in Dataset information.
- When the Dataset management pop-up window appears, select a dataset to connect from the drop-down menu for each item.
- Click the [OK] button.
- Connected dataset information is displayed in Dataset information, and the [Learning management] button is activated.
Proceed with learning
The following describes how to proceed with dataset learning.
Learning cannot be stopped while learning is in progress. Proceed with learning after checking whether the service to be trained is correct and whether the data set to be trained is properly connected.
- Click the environment you are using in the Region menu and Platform menu of the NAVER Cloud Platform console.
- Click the Services > AI Services > AiTEMS menus in that order.
- Click the Service menu.
- Click the desired service to proceed with learning from the list of services.
- Click the [Learning management] button.
- The [Learning management] button is enabled only when a dataset is connected to the service. If the button is disabled, connect the dataset first.
- When the learning management pop-up window appears, set the learning information and click the [Start learning] button.
Recommendation types: select a referral type
- Personalized recommendation: based on past history, it provides data on recommended items that each individual might like
- Related item recommendation: based on the item's related history, it provides recommended item data that each individual might like
- Popular item recommendation: provides data of popular recommended items by ranking
Set HPO: set options related to model creation
- If HPO (Hyperparameter Optimization) is not set, it learns using the optimized algorithm inside AiTEMS. Set the HPO to achieve the optimal learning model.
NoteThe Popular item recommendation type requires HPO to be set, and the group_column and max_group_recommend_top options can only be set when the Popular item recommendation type is selected. The details on setting HPO parameters for each recommendation type are as follows.
Recommendation types Parameter name Description Default value Available range Personalized recommendation and related item recommendation session_max_timestamp Maximum idle time between actions to be regarded as the same sessions (second) 1800 (30 minutes) 5 or more item_top_n The maximum number of products to be used for recommendation, used in order of popularity 20,000 1,000~500,000 min_item_cut The minimum number of conditions for appearing 5 1 or more min_session_length The minimum number of actions within a session 2 2 or more max_user_recommend_top The maximum number of recommended products per user 100 1~1,000 max_relate_recommend_top The maximum number of related recommended products per product 100 1~1,000 remove_history_item Removal status of a product that already has a history of recommending personalized products False - Popular item recommendation group_column List of group columns extracted by popularity - - target_column Target columns to be extracted by popularity - - max_group_recommend_top The maximum number of popular products per group 100 1~1,000 Learning description: enter the description of learning
- From the notification pop-up window, click the [OK] button.
- The learning progress can be checked in the status of learning information.
Test learning results
You can test the results to see if the dataset has been trained correctly.
The following describes how to conduct a test.
- Click the environment you are using in the Region menu and Platform menu of the NAVER Cloud Platform console.
- Click the Services > AI Services > AiTEMS menus in that order.
- Click the Service menu.
- Click the desired service to test results from the list of services.
- Click the [Test recommendation results] button.
- When the pop-up window to test recommendation results appears, select the learning you want to test.
- Enter the required information according to the learning you've chosen.
- Target ID: if you are testing personalized or related recommendation results, enter the following information
- Personalized recommendation: enter USER_ID to look up in learning results
- Related items recommendation: enter the ITEM_ID to look up in the learning results
- Column: if you are testing popular item recommendation results, enter data to look up in learning results.
- Target ID: if you are testing personalized or related recommendation results, enter the following information
- Click the [Search] button.
- You can check the test results in Results at the bottom of the pop-up window.
Change settings of cyclic learning
Cyclic learning allows you to modify the interval at which data is updated, and you can stop or restart learning by changing the learning settings state.
Edit update interval
The update interval defaults to 12 hours, and you can edit the update interval to 1/3/6/12 hours.
The following describes how to edit update interval.
- Click the environment you are using in the Region menu and Platform menu of the NAVER Cloud Platform console.
- Click the Services > AI Services > AiTEMS menus in that order.
- Click the Service menu.
- Click the desired service to edit update interval from the list of services.
- Click the [Edit] button for update interval.
- After selecting the desired update interval to edit, click the [Edit] button.
Change learning settings state
If learning is in progress, you can stop learning by changing it to Unset, and if learning is stopped, you can change it to Set to continue learning.
If you change the learning setting status to Unset, the data is no longer updated, so it cannot be reflected in real-time recommendation results. If you need to reflect real-time recommendation results, be careful not to change to an unset state.
The following describes how to change the learning settings status.
- Click the environment you are using in the Region menu and Platform menu of the NAVER Cloud Platform console.
- Click the Services > AI Services > AiTEMS menus in that order.
- Click the Service menu.
- Click the desired service to change learning settings status from the list of services.
- Click
or
of learning settings status
indicates unset state, click to change to set state.
indicates set state, click to change to unset state.
- If you change the setting status from Unset to Set, when the pop-up window for learning management appears, set the learning information to proceed and click the [Start learning] button.
- For more information on setting learning information, see Proceed with learning.
- The learning settings state changes to Set, and cyclic learning proceeds again.
- If you change the setting status from Set to Unset, check the contents of the notification pop-up window and click the [OK] button.
- The learning setting status is changed to Unset, and cyclic learning is stopped.