Back to projects

Copenhagen Cycling Ecosystem Analysis (Freelanced)

Scraped and analyzed Copenhagen's bicycle retail ecosystem end to end, from Google Maps listings to underserved-market clustering.

Data AnalystData AnalysisMachine LearningWeb ScrapingVisualization

An end-to-end data analysis pipeline studying Copenhagen's bicycle retail and infrastructure ecosystem: Selenium scrapes 311 bicycle-related business listings (name, category, rating, address) from Google Maps, regex extracts Danish zip codes from addresses, and the data is merged with official population, income, transport, and bike-lane statistics across 17 zip codes. Interactive Folium maps (shop density heatmaps, category marker clusters, bike-lane coverage, Haversine-based same-category-neighbor proximity) surface the geospatial picture, K-Means clustering on population-per-shop ratio flags underserved zip codes as concrete market-expansion targets, and linear regression models usage and sales trends across income classes over multiple years. A small Dash app makes the shop dataset interactively explorable.

Combined statistical overview: trip patterns by income class, trip length, bike-lane density, thefts vs. sales trends, and bicycle-type popularity over time
Most underserved zip codes by population-per-shop ratio (zip 2000: about 5,091 people per shop)
Shop count vs. population by zip code: urban hubs dominate both, but retail presence doesn't map perfectly to population density
Rating distribution by category: bicycle stores (164 listings, 4.30 avg) and repair shops (54 listings, 4.37 avg) are the most consistently well-rated

What I built

  • Scraped 311 bicycle-related business listings (name, category, rating, address, About attributes) from Google Maps using Selenium.
  • Extracted Danish zip codes from addresses via regex and merged the scraped data with official population, income, transport, and bike-lane statistics across 17 zip codes.
  • Built interactive Folium maps: shop density heatmaps, category marker clusters, regional bike-lane coverage, and Haversine-distance same-category-neighbor proximity maps.
  • Applied K-Means clustering on population-per-shop ratio to flag underserved zip codes (zip 2000: about 5,091 people per shop) as concrete market-expansion targets.
  • Modeled bicycle usage and sales trends across income classes and time with linear regression on multi-year government transport survey data.
  • Built a Dash app for interactively exploring the shop dataset.

Tech stack

Data AnalystData AnalysisMachine LearningWeb ScrapingVisualization

Key results

311

Shops scraped

17

Zip codes analyzed

~5,091 people/shop

Most underserved zip

Repair shops (4.37 avg)

Top-rated category

Back to all projects