Create a behavioral dataset
You can create a behavioral dataset to filter audiences by behavior or create an AIPL dataset from a behavioral dataset.
Behaviors include user actions related to brands, products, and stores, such as searching, browsing, adding to favorites, adding to a cart, purchasing, liking, commenting, forwarding, and visiting a store. For behavioral datasets used for repurchase prediction and product recommendation, the behaviors must focus on product purchases.
Prerequisites
The behavioral data table is stored in a data source. For a sample format, see Behavioral dataset sample.
If the behavioral dataset will be used for repurchase prediction or product recommendation, the data source that stores the behavioral data table must be an ADB 3.0 data source.
The data source is connected to Quick Audience. For more information, see Create a data source or Grant permissions on a data source table.
Procedure
Go to Workspace > Configuration Management > Data Center > Dataset to open the dataset management page.
In the upper-right corner, click New > Dataset > Behavioral Dataset to open the behavioral dataset configuration page.
Click Select Data Table. In the dialog box that appears, select a data source and a data table, and then click Confirm.
Click Next.
Configure mappings.
Set the mapping fields for the behavioral dataset. These fields include user identity, user identity type, behavior time, behavior type, behavior count, behavior amount, behavior channel, object attribute, and object attribute value.
The user identity is the unique identifier for a user in the dataset.
The user identity type is the ID type that corresponds to the user identity. This becomes the default ID type when you select an audience.
The supported ID types are OneID, UnionID, mobile phone number, mailbox, Taobao ID, Taobao nickname, Taobao OUID, Alipay ID, Weibo ID, International Mobile Equipment Identity (IMEI), identifier for advertisers (IDFA), International Mobile Subscriber Identity (IMSI), Open Anonymous Device Identifier (OAID), MAC address, and OpenID.
After you purchase a repurchase prediction resource plan or a product recommendation resource plan, you must set the mapping field for the object attribute value ID when you create a behavioral dataset for Brand High Potential Forecast or Product Matching Recommendation.
Add dimension tables: Add dimension tables for channel, behavior type, and object attribute to use for behavior filtering.
Select one or more object attribute names, and set the corresponding dimension tables and fields.
For multiple object attributes, if you select All attributes in one dimension table, you can add a single dimension table. This allows data developers to process only one dimension table that contains all object attributes. If you select One dimension table for each attribute, you must add a separate dimension table for each object attribute.
As shown in the following figure, when you filter by behavior or create an AIPL dataset from the behavioral dataset, you must first select an object attribute name. Then, select a value from the object attribute field of the dimension table.
(Optional) Add other ID types: If the selected data table contains other user ID columns, you can add these user IDs to the dataset. When you generate an audience from this model, you can select multiple ID types for push notifications.
ImportantSet an ID type for the new ID field. This ensures that you can select multiple ID types when you push an audience package or a standard AIPL model that is generated from this behavioral dataset to Data Bank.
Click Finish. In the dialog box that appears, enter a dataset name and select a save location. Then, click Confirm to create the behavioral dataset.
You are redirected to the dataset management page, where the new behavioral dataset appears in the dataset list. For more information about related management operations, see Manage a behavioral dataset.