All Products
Document Center

Artificial Intelligence Recommendation:Preparations

Last Updated:Mar 07, 2023


Welcome to Artificial Intelligence Recommendation (AIRec). You will learn how to get started with ease, quickly find help information, and achieve a better user experience.

This topic describes the preparations required to deploy an AIRec instance and the main features of the AIRec console. After you complete the preparations, you must select an industry template and a scenario template that suit your business requirements. You can select a method to upload data based on your data. This way, data quality is ensured when you run recommendation tests and deploy online services.


  • Logic of event tracking on behavior data: the logic that is followed to collect user behavior data that meets AIRec specifications. The AIRec algorithm uses behavior data to learn user preferences.

  • Boot data: the existing historical data that is prepared before an AIRec instance is started. The data can be used to quickly start the model of AIRec algorithm for recommendations.

  • Real-time data: the incremental data that is uploaded by using a software development kit (SDK) after an AIRec instance is started. ------------------

  • DataWorks and MaxCompute: two cloud-native big data services provided by Alibaba Cloud. They can be used together with AIRec.

1. Prepare data based on data specifications

The data specifications vary based on different industries. To prepare data, follow the instructions described in specific topics.

If you want to test features or API operations first, you can upload the test data.

Test data

AIRec provides data that can be used for tests. You can obtain the data from the following topic:

Download test data

2. Check whether the logic of event tracking on behavior data meets specifications

Data about exposure and click behavior is required for event tracking. The number of items that are exposed must be larger than the number of items that are clicked. For more information, see Use event tracking.

If you cannot provide exposure data, see the 15. What can I do if no exposure data or trace information can be provided based on the instrumentation logic of my application? section in the FAQ related to data access topic.

3. Determine a suitable method to upload data

Note: The method to upload data consists of two steps: push boot data and push real-time data.

AIRec provides multiple methods to push boot data and real-time data. You can select a method based on your data forms. When data changes, make sure that the changes can be synchronized to AIRec in real time no matter which method is used.

You can choose a suitable method in each step based on your business requirements.



Push boot data

Without historical data:

Quickly start an instance.

With historical data:

Upload historical data by using MaxCompute and start an instance.

Push real-time data

Use server SDKs:

Push and update user data, item data, and behavior data by using server SDKs.

The following section describes the procedure for each method:

Quickly start an instance

Use server SDKs to upload data

Use server SDKs to start an instance

Overview: Use server SDKs to upload user data, item data, and behavior data.Use server SDKs to upload all the data to quickly start an instance

Use historical data to start an instance

Use historical data to start an instance

Overview: Upload user data and item data, as well as the behavior data that was generated within the last one to two weeks, to MaxCompute, and validate the data. We recommend that you use DataWorks to upload the data. Then, use server SDKs to synchronize the changes of user data and item data and to upload behavior data that is generated in real time.Historical data