Feature selector

Feature Selector

What Is a Feature Selector

A feature selector filters the users and events that have been collected by setting a series of conditions, so that you can obtain the final subset of data you need for use and analysis. Many features in Jiguang Growth, such as user segments, smart labels, and in-app marketing scores, use feature selectors.

A feature selector is similar to a where condition in SQL.
When it is applied to users, the where condition acts on the user table. When it is applied to events, the where condition acts on the event table.
Feature Selector Categories
Feature selectors are divided into event-based feature selectors and non-event feature selectors.
Event-Based Feature Selector
An event-based feature selector can act on either the event table or the user table, and can be used to view events or users that meet the selected conditions. Event attributes and user attributes are reported together with events. When conditions are set on event attributes or user attributes, they act on the event table to obtain matching events, and then obtain the corresponding users through the event selection conditions. When conditions are set on user labels, user marketing scores, or user segments, they act on the user table and directly obtain matching users.
Non-Event Feature Selector
A non-event feature selector is also a user feature selector. Its conditions act on the user table and only look at users that meet the selected conditions. Conditions can be set on user attributes, user labels, user marketing scores, and user segments to obtain matching users.
Feature Filter Types
Whether you are filtering different users or different events, each filter feature can be classified by its data type.
String Filtering
When the filtered data is a string, refer to the following table for how to use each judgment type:

Judgment Type	SQL Expression	Description
Equal to	=	Exact match. The user or event is selected only when the selected attribute equals the entered judgment value. When multiple judgment values are entered, they have an "OR" relationship.
Not equal to	!=	Exact match. The user or event is selected only when the selected attribute is not equal to the entered judgment value. When multiple judgment values are entered, they have an "AND" relationship.
Contains	LIKE "%judgment value%"	Fuzzy match. The user or event is selected only when the selected attribute contains the entered judgment value. When multiple judgment values are entered, they have an "OR" relationship.
Does not contain	NOT LIKE "%judgment value%"	Fuzzy match. The user or event is selected only when the selected attribute does not contain any of the entered judgment values. When multiple judgment values are entered, they have an "AND" relationship.
Has value	IS NOT NULL AND!=""	The user or event can be selected only when the selected attribute value is not `NULL` and is not empty.
No value	IS NULL OR =""	The user or event is selected only when the selected attribute value is `NULL` or empty.

Numeric Filtering

When the filtered data is numeric, refer to the following table for how to use each judgment type:

Judgment Type	SQL Expression	Description
Equal to	=	Same as string filtering.
Not equal to	!=	Same as string filtering.
Less than	<	The user or event is selected only when the selected attribute value is less than the judgment value.
Greater than	>	The user or event is selected only when the selected attribute value is greater than the judgment value.
Range	between judgment value 1 and judgment value 2	The user or event is selected only when the selected attribute value falls within the closed interval formed by the judgment values. For example, if the judgment values are set to 10 and 100, the selected attribute value must satisfy [10,100].
Has value	IS NOT NULL AND!=""	Same as string filtering.
No value	IS NULL OR =""	Same as string filtering.

Boolean Filtering

When the filtered data is Boolean, refer to the following table for how to use each judgment type:

Judgment Type	SQL Expression	Description
True	=1	The user or event is selected only when the selected attribute value is true.
False	=0	The user or event is selected only when the selected attribute value is false.
Has value	IS NOT NULL AND!=""	Same as string filtering.
No value	IS NULL OR =""	Same as string filtering.

Date-Time Filtering

When the filtered data is date-time data, refer to the following table for how to use each judgment type:

Judgment Type	SQL Expression	Description
Absolute time	between start time and end time	The user or event is selected only when the selected attribute falls between the start time and end time.
Relative to current time		The user or event is selected only when the selected attribute falls within a time range relative to the current time. For example, if today is the 17th and the condition is within the past 7 days, the time range is from the 11th to the 17th, including both the 11th and the 17th.
Has value	IS NOT NULL AND!=""	Same as string filtering.
No value	IS NULL OR =""	Same as string filtering.

Collection Filtering

When the filtered data is a collection, refer to the following table for how to use each judgment type:

Judgment Type	SQL Expression	Description
Contains	="judgment value"	Exact match. The user or event is selected only when the selected attribute equals any of the entered judgment values. When multiple judgment values are entered, they have an "OR" relationship.
Does not contain	!="judgment value"	Exact match. The user or event is selected only when the selected attribute is not equal to all entered judgment values. When multiple judgment values are entered, they have an "AND" relationship.
Has value	IS NOT NULL AND!=""	The selected attribute value can be selected only when at least one element exists and satisfies the non-empty string condition.
No value	IS NULL OR =""	The selected attribute value is selected only when no element exists or all elements are empty strings.

Data Value Rules

When a feature selector is used in different feature scenarios, the value rules vary. In general, there are two types:

At that time: takes the feature value at the time when the event was reported.
Latest: takes the current latest feature value.

Event-based feature filtering:

Feature Scenario	Event Attribute	User Attribute	User Segment	User Label	User Marketing Score
Create virtual event	At that time	At that time	/	/	/
Create combined event	At that time	At that time	/	/	/
Create operation plan	At that time	At that time	/	/	/
Create user label	At that time	At that time	Latest	/	Latest
Create user segment	At that time	At that time	/	Latest	Latest
Create user marketing score	At that time	At that time	Latest	Latest	/
Create behavior analysis	At that time	At that time	Latest	Latest	Latest

Non-event feature filtering:

Feature Scenario	User Attribute	User Segment	User Label	User Marketing Score
Create user segment	Latest	/	Latest	Latest
Create marketing score	Latest	Latest	Latest	/
Create user label	Latest	Latest	/	Latest

What Is a Feature Selector