Skip to navigation
Kumo Studio Guide (Coming Soon)Preliminary DocumentationTrain ModelPredictive Query

Task Types

View as Markdown
Prerelease documentation

This Kumo Studio information is prerelease and is not yet supported. Stay tuned for the Docker and Kubernetes releases.

The task type defines the kind of prediction your Predictive Query (PQL) produces. Kumo determines it automatically from the structure of your query. The task type affects:

  • The commands and operators used in the query.

  • The available evaluation metrics.

  • Explainable AI (XAI) options.

Common Task Types

Task TypeOutputPQL Example
RegressionContinuous real numberPREDICT customers.age FOR EACH customers.customer_id
Binary ClassificationTrue or FalsePREDICT fraud_reports.is_fraud FOR EACH transactions.id WHERE transactions.type = "bank transfer"
Multiclass + Multilabel ClassificationClass labelPREDICT FIRST(purchases.type, 0, 7) FOR EACH users.user_id
Link PredictionList of itemsPREDICT LIST_DISTINCT(transactions.article_id, 0, 7) RANK TOP 10 FOR EACH customers.customer_id

Examples

Regression

A regression task predicts a continuous value. For example, predicting the total amount of purchases per customer over the next 30 days:

PQL
PREDICT SUM(TRANSACTIONS.PRICE, 0, 30, days)
FOR EACH CUSTOMERS.CUSTOMER_ID

Kumo recognizes this as a regression task since it uses the SUM() operator without a boolean condition.

Binary Classification

A binary classification task predicts a true/false outcome. For example, predicting which customers will make no transactions in the next 30 days:

PQL
PREDICT COUNT(TRANSACTIONS.*, 0, 30, days) = 0
FOR EACH CUSTOMERS.CUSTOMER_ID

  • The = operator makes this a classification task.

  • Using > instead changes the positive label of the prediction.

For more examples across task types, see Predictive Query Examples.