How do I cluster by features in CartoDB?

Question

I have a set of points for a number of buildings in NYC. I would like to group those points into clusters per their respective neighborhood (ex: Chelsea, Upper East Side etc.). Each of my points have a designated neighborhood.

Is there a way to tweak the CSS to make it happen? I don't quite understand how the Wizard is currently clustering my points:

    /** cluster visualization */

#data{
  marker-width: 12;
  marker-fill: #FD8D3C;
  marker-line-width: 1.5;
  marker-fill-opacity: 1;
  marker-line-opacity: 1;
  marker-line-color: #fff;
  marker-allow-overlap: true;

  [src = 'bucketE'] {
    marker-line-width: 5;
    marker-width: 24;
  } 

  [src = 'bucketD'] {
    marker-line-width: 5;
    marker-width: 34;
  } 

  [src = 'bucketC'] {
    marker-line-width: 5;
    marker-width: 44;
  } 

  [src = 'bucketB'] {
    marker-line-width: 5;
    marker-width: 54;
  } 

  [src = 'bucketA'] {
    marker-line-width: 5;
    marker-width: 64;
  } 
}

#data::labels { 
  text-size: 0; 
  text-fill: #fff; 
  text-opacity: 0.8;
  text-name: [points_count]; 
  text-face-name: 'DejaVu Sans Book'; 
  text-halo-fill: #FFF; 
  text-halo-radius: 0; 

  [src = 'bucketE'] {
    text-size: 12;
    text-halo-radius: 0.5;
  }

  [src = 'bucketD'] {
    text-size: 17;
    text-halo-radius: 0.5;
  }

  [src = 'bucketC'] {
    text-size: 22;
    text-halo-radius: 0.5;
  }

  [src = 'bucketB'] {
    text-size: 27;
    text-halo-radius: 0.5;
  }

  [src = 'bucketA'] {
    text-size: 32;
    text-halo-radius: 0.5;
  }

  text-allow-overlap: true;

  [zoom>4]{ text-size: 16; }
  [points_count = 1]{ text-size: 0; }
}

Answer 1

I'm not sure if you can do it with pure styling, but by manipulating your tables some you can likely get your desired result.

I would find a shapefile of NYC neighborhoods and upload that to Cartodb, then merge tables with a spatial join. Likely just doing count, to get the number of features, though you could also sum on a property.

That will give you a table of the neighborhoods, with the number of your features as a column. Personally I'd just display the polygons with a label of the count of the features. But if you really wanted it as a circle you could take the centroid of the neighborhood, populating it with ST_Centroid(the_geom) from your table (easiest would probably be to create a new table). Then you can style that as you like.