Online coding and Python classes in Runcorn

The learner downloads a rectangle covering Runcorn from the OpenStreetMap API in eight tiles and keeps every place tagged as a shop, café, restaurant or takeaway: 272 of them. Their positions are converted to metres. DBSCAN then needs two settings. One is the minimum crowd, set here at five: a place with at least five places, itself included, within reach counts as a core point. The other is the radius that defines "within reach". Clusters are chains of core points and their neighbours; anything left over is noise.

Two clusters dominate at every radius. The larger one, whose most common tagged street address is Church Street, holds between 100 and 127 places; the second, where Forest Walk is the most common address, holds between 54 and 68. The program found them with no idea of how many groups to expect, which is exactly what DBSCAN is for. A method such as k-means, by contrast, has to be told the number of clusters in advance.

Everything smaller depends on the radius. Widening it from 60 m to 150 m pulls more lone shops into small clusters, so the number of clusters climbs from 5 to 8 and the noise falls from 34.9% to 23.5%. Widen it again to 250 m and small clusters start merging into their neighbours, so the count drops back to six. There is no single "correct" radius. The honest summary is that two shopping areas are robust and the rest are sensitive to the setting, which is the kind of statement a careful analyst makes and a careless one skips.

Knowing which patterns are solid and which are fragile helps Runcorn teenagers use AI analysis wisely, and that is worth learning Python for in 2026. The longer argument is in why teenagers should still learn to code in 2026.