> For the complete documentation index, see [llms.txt](https://docs.carto.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.carto.com/data-and-analysis/analytics-toolbox-for-redshift/sql-reference/clustering.md).

# clustering

This module contains functions that perform clustering on geographies.

## CREATE\_CLUSTERDBSCAN <a href="#create_clusterdbscan" id="create_clusterdbscan"></a>

```sql
CREATE_CLUSTERDBSCAN(input, output_table, geom_column, epsilon, min_points [, partition_column])
```

**Description**

Takes a set of points as input and groups them into clusters using the DBSCAN algorithm, writing the result to a new table.

DBSCAN groups together points lying in dense neighborhoods and labels the rest as noise. Unlike k-means it does not require the number of clusters up front, it finds clusters of arbitrary shape, and it does not force every point into a cluster.

**Input parameters**

* `input`: `VARCHAR` name of the table or literal SQL query to be clustered.
* `output_table`: `VARCHAR(MAX)` qualified name of the output table, e.g. `<my-schema>.<my-output-table>`. It is replaced if it already exists.
* `geom_column`: `VARCHAR` name of the `POINT` column to be clustered, in SRID 4326 or 0.
* `epsilon`: `FLOAT8` the search radius in meters.
* `min_points`: `INT` the minimum number of points, counting the point itself, that form a dense neighborhood.
* `partition_column` (optional): `VARCHAR` name of a column to cluster within, `NULL` values forming their own group. If omitted the whole input is one set.

**Output**

The output table contains all the columns of `input` plus `cluster_id` and `pt_type`.

`cluster_id` is a zero-based cluster index, restarting at zero in each partition. It is `NULL` for any point that is not in a cluster.

`pt_type` is the role of the point. A **core** point has at least `min_points` points within `epsilon` of it, counting itself, and core points within `epsilon` of each other belong to the same cluster. A **border** point is not a core point but lies within `epsilon` of one, so it joins that cluster without connecting it to any other. A **noise** point is neither. A **skipped** point was never clustered because its geometry was `NULL`, which is absent input rather than a result.

{% hint style="info" %}
**info**

Cluster assignments match `sklearn.cluster.DBSCAN` with `metric='haversine'` for the same `epsilon` and `min_points`. Where DBSCAN is inherently ambiguous — a border point reachable from two clusters — this implementation always picks the cluster with the lowest canonical label, so results are deterministic and reproducible across runs.
{% endhint %}

{% hint style="warning" %}
**warning**

Only `POINT` geometries are supported. The input must not already have columns named `cluster_id`, `pt_type` or `__carto_idx`, so the output of one call cannot be fed straight back into another.

Runtime is driven by point *density* rather than row count: the cost grows with the number of points within `epsilon` of each other, so a large radius over a tightly packed area is the expensive case.
{% endhint %}

**Examples**

{% code overflow="wrap" lineNumbers="true" %}

```sql
CALL carto.CREATE_CLUSTERDBSCAN('<my-schema>.<my-table>', '<my-schema>.<my-output-table>', 'geom', 100, 5);
-- The table `<my-schema>.<my-output-table>` will be created adding the columns
-- cluster_id and pt_type to those in `<my-schema>.<my-table>`.
```

{% endcode %}

{% code overflow="wrap" lineNumbers="true" %}

```sql
CALL carto.CREATE_CLUSTERDBSCAN('SELECT * FROM <my-schema>.<my-table>', '<my-schema>.<my-output-table>', 'geom', 100, 5);
-- The table `<my-schema>.<my-output-table>` will be created adding the columns
-- cluster_id and pt_type to those returned in the input query.
```

{% endcode %}

{% code overflow="wrap" lineNumbers="true" %}

```sql
CALL carto.CREATE_CLUSTERDBSCAN('<my-schema>.<my-table>', '<my-schema>.<my-output-table>', 'geom', 25, 3, 'store_id');
-- Points are clustered independently for each store_id, in a single call.
```

{% endcode %}

{% code overflow="wrap" lineNumbers="true" %}

```sql
-- Keep only the clustered points and count them per cluster
SELECT cluster_id, COUNT(*) AS points
FROM <my-schema>.<my-output-table>
WHERE cluster_id IS NOT NULL
GROUP BY cluster_id
ORDER BY points DESC;
```

{% endcode %}

## CREATE\_CLUSTERKMEANS <a href="#create_clusterkmeans" id="create_clusterkmeans"></a>

```sql
CREATE_CLUSTERKMEANS(input, output_table, geom_column, number_of_clusters)
```

**Description**

Takes a set of points as input and partitions them into clusters using the k-means algorithm. Creates a new table with the same columns as `input` plus a `cluster_id` column with the cluster index for each of the input features.

**Input parameters**

* `input`: `VARCHAR` name of the table or literal SQL query to be clustered.
* `output_table`: `VARCHAR(MAX)` qualified name of the output table, e.g. `<my-schema>.<my-output-table>`. The process will fail if the table already exists.
* `geom_column`: `VARCHAR` name of the column to be clusterd.
* `number_of_clusters`: `INT` number of clusters that will be generated.

{% hint style="warning" %}
**warning**

Keep in mid that due to some restrictions in the Redshift `VARCHAR` size, the maximum number of features (points) allow to be clustered is around 2500.
{% endhint %}

**Examples**

{% code overflow="wrap" lineNumbers="true" %}

```sql
CALL carto.CREATE_CLUSTERKMEANS('<my-schema>.<my-table>', '<my-schema>.<my-output-table>', 'geom', 5);
-- The table `<my-schema>.<my-output-table>` will be created
-- adding the column cluster_id to those in `<my-schema>.<my-table>`.
```

{% endcode %}

{% code overflow="wrap" lineNumbers="true" %}

```sql
CALL carto.CREATE_CLUSTERKMEANS('SELECT * FROM <my-schema>.<my-table>', '<my-schema>.<my-output-table>', 'geom', 5);
-- The table `<my-schema>.<my-output-table>` will be created
-- adding the column cluster_id to those returned in the input query.
```

{% endcode %}

## ST\_CLUSTERKMEANS <a href="#st_clusterkmeans" id="st_clusterkmeans"></a>

```sql
ST_CLUSTERKMEANS(geog [, numberOfClusters])
```

**Description**

Takes a set of points as input and partitions them into clusters using the k-means algorithm. Returns an array of tuples with the cluster index for each of the input features and the input geometry.

**Input parameters**

* `geog`: `GEOMETRY` points to be clustered.
* `numberOfClusters` (optional): `INT` number of clusters that will be generated. It defaults to the square root of half the number of points (`sqrt(<NUMBER OF POINTS>/2)`). The output number of cluster cannot be greater to the number of distinct points of the `geog`.

**Return type**

`SUPER`: containing objects with `cluster` as the cluster id and `geom` as the geometry in GeoJSON format.

**Examples**

{% code overflow="wrap" lineNumbers="true" %}

```sql
SELECT carto.ST_CLUSTERKMEANS(ST_GEOMFROMTEXT('MULTIPOINT ((0 0), (0 1), (5 0), (1 0))'));
-- {"cluster":0,"geom":{"type":"Point","coordinates":[0.0,0.0]}}
-- {"cluster":0,"geom":{"type":"Point","coordinates":[0.0,1.0]}}
-- {"cluster":0,"geom":{"type":"Point","coordinates":[5.0,0.0]}}
-- {"cluster":0,"geom":{"type":"Point","coordinates":[1.0,0.0]}}
```

{% endcode %}

{% code overflow="wrap" lineNumbers="true" %}

```sql
SELECT carto.ST_CLUSTERKMEANS(ST_GEOMFROMTEXT('MULTIPOINT ((0 0), (0 1), (5 0), (1 0))'), 2);
-- {"cluster":0,"geom":{"type":"Point","coordinates":[0.0,0.0]}}
-- {"cluster":0,"geom":{"type":"Point","coordinates":[0.0,1.0]}}
-- {"cluster":1,"geom":{"type":"Point","coordinates":[5.0,0.0]}}
-- {"cluster":0,"geom":{"type":"Point","coordinates":[1.0,0.0]}}
```

{% endcode %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.carto.com/data-and-analysis/analytics-toolbox-for-redshift/sql-reference/clustering.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.