Select Page

GIS Copilot Demo: K-Mean clustering and visualization

Task

Perform K-mean clustering for the points and visualize the cluster results.

Dataset

Any point data

Result

Geoprocessing workflow

Generated code: 

  1. # This script performs spatial clustering of a point dataset using K-means and visualizes the resulting clusters.
  2. # Steps:
  3. # 1) Reproject (skipped because input is already in a projected CRS in meters: EPSG:3857).
  4. # 2) Add Geometry Attributes: store X and Y coordinates using ‘qgis:exportaddgeometrycolumns’.
  5. # 3) K-Means Clustering: run ‘native:kmeansclustering’ with a specified K to create ‘CLUSTER_ID’ and cluster size field.
  6. # 4) Calculate Cluster Centroids: use ‘native:meancoordinates’ grouped by ‘CLUSTER_ID’.
  7. # 5) Visualize Clusters: apply unique value symbology by ‘CLUSTER_ID’, symbolize cluster centroids, and create convex hulls per cluster (optional).
  8. # Notes:
  9. # – All outputs are saved to the specified output directory with unique filenames to avoid overwrite conflicts.
  10. # – Each processing step is followed by a short waiting period to ensure file creation before subsequent steps.
  11. # – Uses only required PyQGIS and processing tools; geopandas/numpy are not necessary here.
  13. import os
  14. import time
  15. import random
  16. import processing
  17. from qgis.core import (
  18.     QgsVectorLayer,
  19.     QgsProject,
  20.     QgsSymbol,
  21.     QgsRendererCategory,
  22.     QgsCategorizedSymbolRenderer,
  23.     QgsSingleSymbolRenderer
  24. )
  25. from PyQt5.QtGui import QColor
  28. def perform_spatial_clustering():
  29.     # Inputs
  30.     input_path = r“C:\Users\17036\Downloads\test\Hazardous_Waste_Sites\HW_Sites.shp”
  31.     output_dir = r“C:\Users\17036\AppData\Roaming\QGIS\QGIS3\profiles\default\python\plugins\SpatialAnalysisAgent-master\Default_workspace”
  32.     k_clusters = 5  # Specify number of clusters
  34.     # Helper to generate a unique shapefile path in output directory
  35.     def unique_path(base_name, ext=“.shp”):
  36.         base = os.path.join(output_dir, base_name + ext)
  37.         if not os.path.exists(base):
  38.             return base
  39.         i = 1
  40.         while True:
  41.             candidate = os.path.join(output_dir, f“{base_name}_{i}{ext}”)
  42.             if not os.path.exists(candidate):
  43.                 return candidate
  44.             i += 1
  46.     # Load input point layer
  47.     input_layer = QgsVectorLayer(input_path, “HW_Sites”, “ogr”)
  48.     # Step 1: Reproject (skipped) – Input CRS is EPSG:3857 (meters), appropriate for K-means Euclidean distance
  49.     reprojected_layer = input_layer
  50.     reprojected_path = input_path
  52.     # Step 2: Add geometry attributes (X and Y)
  53.     geom_out_path = unique_path(“HW_Sites_with_geom”)
  54.     add_geom_params = {
  55.         ‘INPUT’: reprojected_path,
  56.         ‘CALC_METHOD’: 0,  # Layer CRS
  57.         ‘OUTPUT’: geom_out_path
  58.     }
  59.     add_geom_result = processing.run(‘qgis:exportaddgeometrycolumns’, add_geom_params)
  60.     time.sleep(1)
  61.     geom_layer = QgsVectorLayer(add_geom_result[‘OUTPUT’], ‘HW_Sites_with_geom’, ‘ogr’)
  62.     QgsProject.instance().addMapLayer(geom_layer)
  64.     # Step 3: K-means clustering on point geometry
  65.     clusters_out_path = unique_path(“HW_Sites_clusters”)
  66.     kmeans_params = {
  67.         ‘INPUT’: geom_layer,
  68.         ‘CLUSTERS’: k_clusters,
  69.         ‘OUTPUT’: clusters_out_path,
  70.         ‘FIELD_NAME’: ‘CLUSTER_ID’,  # 10 char limit (shapefile)
  71.         ‘SIZE_FIELD_NAME’: ‘CLS_SIZE’  # <= 10 chars
  72.     }
  73.     kmeans_result = processing.run(‘native:kmeansclustering’, kmeans_params)
  74.     time.sleep(1)
  75.     clusters_layer = QgsVectorLayer(kmeans_result[‘OUTPUT’], ‘HW_Sites_clusters’, ‘ogr’)
  76.     QgsProject.instance().addMapLayer(clusters_layer)
  78.     # Step 4: Calculate mean coordinates (centroids) grouped by cluster_id
  79.     fields = clusters_layer.fields()  # Access fields before using any field from the layer
  80.     centroids_out_path = unique_path(“HW_Sites_cluster_centroids”)
  81.     meancoords_params = {
  82.         ‘INPUT’: clusters_layer,
  83.         ‘WEIGHT’: None,
  84.         ‘UID’: ‘CLUSTER_ID’,  # group by cluster id
  85.         ‘OUTPUT’: centroids_out_path
  86.     }
  87.     meancoords_result = processing.run(‘native:meancoordinates’, meancoords_params)
  88.     time.sleep(1)
  89.     centroids_layer = QgsVectorLayer(meancoords_result[‘OUTPUT’], ‘HW_Sites_cluster_centroids’, ‘ogr’)
  90.     QgsProject.instance().addMapLayer(centroids_layer)
  92.     # Step 5 (optional): Convex hull per cluster to depict spatial extents
  93.     hulls_out_path = unique_path(“HW_Sites_cluster_hulls”)
  94.     hull_params = {
  95.         ‘INPUT’: clusters_layer,
  96.         ‘FIELD’: ‘CLUSTER_ID’,
  97.         ‘OUTPUT’: hulls_out_path
  98.     }
  99.     hull_result = processing.run(‘native:convexhull’, hull_params)
  100.     time.sleep(1)
  101.     hulls_layer = QgsVectorLayer(hull_result[‘OUTPUT’], ‘HW_Sites_cluster_hulls’, ‘ogr’)
  102.     QgsProject.instance().addMapLayer(hulls_layer)
  104.     # Visualization: Unique value symbology for clusters (by CLUSTER_ID)
  105.     cluster_field = ‘CLUSTER_ID’
  106.     categories = []
  107.     unique_vals = set(f[cluster_field] for f in clusters_layer.getFeatures())
  108.     for v in sorted(unique_vals):
  109.         sym = QgsSymbol.defaultSymbol(clusters_layer.geometryType())
  110.         sym.setColor(QColor(random.randint(0, 230), random.randint(0, 230), random.randint(0, 230), 220))
  111.         try:
  112.             sym.setSize(3.2)
  113.         except:
  114.             pass
  115.         cat = QgsRendererCategory(v, sym, str(v))
  116.         categories.append(cat)
  117.     cat_renderer = QgsCategorizedSymbolRenderer(cluster_field, categories)
  118.     clusters_layer.setRenderer(cat_renderer)
  119.     clusters_layer.triggerRepaint()
  121.     # Visualization: Centroids (single symbol)
  122.     cent_sym = QgsSymbol.defaultSymbol(centroids_layer.geometryType())
  123.     cent_sym.setColor(QColor(0, 0, 0))
  124.     try:
  125.         cent_sym.setSize(5.0)
  126.     except:
  127.         pass
  128.     centroids_layer.setRenderer(QgsSingleSymbolRenderer(cent_sym))
  129.     centroids_layer.triggerRepaint()
  131.     # Visualization: Cluster hulls (semi-transparent fill)
  132.     hull_sym = QgsSymbol.defaultSymbol(hulls_layer.geometryType())
  133.     hull_sym.setColor(QColor(255, 215, 0, 50))  # light transparent yellow
  134.     hulls_layer.setRenderer(QgsSingleSymbolRenderer(hull_sym))
  135.     hulls_layer.triggerRepaint()
  138. perform_spatial_clustering()