> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/sdgm/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/sdgm/_mcp/server.

# Understanding Predictive Query

## What is a Predictive Query?

A **Predictive Query** is a **declarative syntax** used to define a predictive modeling task in Kumo. It specifies the target variable to predict and the data context for training.

Kumo uses the **Predictive Query Language (PQL)**, a SQL-like language, to automate all major steps of the ML pipeline, including **feature engineering, generating a training table, and training a model**.

## Creating a Predictive Query

To train a model in Kumo:

1. Navigate to **New > Model** from the side menu.

2. On the **Model Training** page, enter a **Model Name** and optional **Description**.

3. Select a **Graph** from your previously created graphs. Once a graph is selected, its structure and linkages appear on the right side for reference.

4. Write your **Predictive Query (PQL)** in the text area.

**Example PQL**

```sql
PREDICT COUNT(transactions.*, 0, 30, days) = 0
FOR EACH customers.customer_id
WHERE COUNT(transactions.*, -90, 0, days) > 0
```

![](/sdgm/_files/nvidia-sdgm.docs.buildwithfern.com/352c2ccab259e4ffee103a9e85cf4aa3dda726b77c30ec3aa211cee36e3a0905/img/pq-entered.png)

## Model Settings

Before training, users can access **advanced model settings** to:

* Configure baseline and run mode.

* Adjust hyperparameters in the model plan.

* Modify training table generation settings.

By default, Kumo optimizes the model automatically, but advanced settings allow customization based on specific needs.

![](/sdgm/_files/nvidia-sdgm.docs.buildwithfern.com/5a0e26aed5ab8015f3962879e433fc08d56fe650b0476d8338345e18c1393bca/img/model-settings.png)