Artificial Intelligence Recommendation:Use historical data to start an instance

MaxCompute is activated and you are familiar with related operations.

3.1. Baseline data

Import baseline user data and item data into MaxCompute. Import behavioral data that was generated within the last one to two weeks into MaxCompute.

3.2. Incremental data

Use server SDKs to upload incremental user data, item data, and behavioral data.

3.3. Use of server SDKs

If you use Artificial Intelligence Recommendation (AIRec) SDK for Java, perform the following operations to use server SDKs:

a. Configure the environment before you can use AIRec SDK for Java. For more information about the configurations, see SDK for Java.

b. Use AIRec SDK for Java to push data. For more information, see Push data.

c. Obtain recommendation results. For more information, see Obtain recommendation results.

For more information about how to use server SDKs for other programming languages, see the related topics that are under the node of the same level as the "SDK for Java" topic.

3.4. Parameters of tables pushed by server SDKs

3.5. Details about data pushing

Push data

4.1. Select an industry template

Note: After you start an instance, you cannot change the selected industry.

4.2. Select a method to start an AIRec instance

Select Historical Data-Based Start.

Note: After you use baseline data in MaxCompute to start an AIRec instance, you no longer need to maintain data in MaxCompute. You can use server SDKs to upload incremental data.

4.3. Configure data sources to start an AIRec instance

Note:

1. Before you configure data sources, you must grant read permissions to AIRec in MaxCompute. For more information, see Grant permissions on offline storage to AIRec.

2. You need to mask the information provided to Alibaba Cloud. Do not include personal information, such as your name, age, and address.

3. If you use MaxCompute to upload historical data, import the three tables at the same time. You are not allowed to separately update one of the three tables.

4.4. Configure real-time data sources

If you need to start an AIRec instance by using data in MaxCompute, you can use server SDKs to push the incremental behavioral data, item data, and user data. For more information, see Push data.

Click Next.

4.5. Configure offline storage

If you enable the offline storage feature, offline data that is used for model training for AIRec is stored in MaxCompute that is activated by using your enterprise account. You must make sure that the offline data is stable and your enterprise account can use and maintain the data.

After you start an instance, you can view the partition information of offline storage tables on the Basic configurations page in the AIRec console.

Note:

1. The name of the offline storage project must be different from the name of the project whose data you used to start the instance.

2. Before you configure offline data sources, you must grant read and write permissions to AIRec in MaxCompute. For more information, see Grant permissions on offline storage to AIRec.

3. We recommend that you do not modify or delete the offline storage project and enable auto-renewal for the project to avoid overdue payments.

4. After you start an instance, you can view the partition information of offline storage tables on the Offline Storage page in the AIRec console.

5. The current version of AIRec allows you to manage offline data by using the MaxCompute client.

The current version of AIRec supports all the operations in MaxCompute. Proceed with caution if you need to perform high-risk operations. For more information, see MaxCompute.

6. Do not perform high-risk operations on offline data, such as adding, deleting, or writing data. These operations may lead to data calculation failures or affect the performance of data calculation.

After you start an instance, you can view the partition information of offline storage tables on the Offline Storage page in the AIRec console.

Note: If you started an instance before April 14 by using historical data and want to restart the instance, the restart operations are different from the previous start operations. Follow the descriptions in this topic to restart the instance.

4.6. Start an instance

In the Instance Start Confirmation message, click OK. It takes about one and a half hours to start the instance.

View instance details

After the instance is started, you can view details about the instance.

4.7. Incremental data

Use server SDKs to upload incremental data. For more information about the operations, see "3.4. Parameters of tables pushed by server SDKs" and "3.5. Details about data pushing".

Note: You must first push item data and user data, and then push behavioral data.

4.8. User experience tests

In the returned result of a user experience test, you can perform one of the following operations: Click, Favorite, Add, and Buy. A behavioral data record is automatically generated in the background each time you perform one of these operations. These data records help you verify the test effect. When you request test results again, the returned results vary based on the operation that you performed. If you want to view details about test results, go to the Update Message Query page.