Get clusters from a GeoDjango model that could be used in a heatmap (eg, using Heatmap.js), including potentially dynamic filters on the model.

gistfile1.py

from django.db import connection, DEFAULT_DB_ALIAS

num_clusters = 6

queryset = model.objects.filter() # Your filters here
queryset = queryset.filter(centroid__isnull=False).only('centroid')

# Get the query for 
sql_model_query, params = queryset.query.get_compiler(DEFAULT_DB_ALIAS).as_sql()

# Requires Postgresql with the kmeans extension 
cluster_query = """
SELECT kmeans AS id, count(*), ST_Centroid(ST_Collect(centroid)) AS centroid
FROM (
    SELECT kmeans(ARRAY[ST_X(centroid), ST_Y(centroid)], %d) OVER (), centroid
    FROM (%s) as centroid
) AS ksub
GROUP BY kmeans
ORDER BY kmeans
""" % (num_clusters, sql_model_query)

counts = []
max_count = 0

# Execute the query
cursor = connection.cursor()
cursor.execute(cluster_query, params)
for cluster in cursor.fetchall():
    # Assemble a heatmap.js-friendly dict
    id, count, centroid = cluster
    max_count = max(max_count, count)
    centroid = GEOSGeometry(centroid)
    counts.append({
        'lat': centroid.y,
        'lon': centroid.x,
        'value': count,
    })

heatmap_data = {
    'max': max_count,
    'data': counts,
}