uft93 / coursera_capstone

I intended to cluster the districts of Munich in this notebook, depending on their most frequent venue types. The district information was obtained through scraping Munich's city website. The pertinent geographic information was obtained through calls to the Google Maps API. The venue information is the courtesy of FourSquare API. This notebook was originally completed for an older iteraton of the final project of IBM Data Science Specialization on Coursera.

Clustering the Districts of Munich

This was my first personal final project for Coursera IBM Data Science specialization. I eventually ended up doing another project for that specialization but decided to keep this one as this has been my longest and toughest project so far. Unsupervised learning always presents a challenge.

The project revolves around using unsupervised learning to cluster the neighborhoods of Munich in a way that makes sense.

My tasks for this project were:

• Scrape the Munich city portal for the district name - postcode pairs
• Make calls to the Google Maps API to create a DataFrame with district names, postcodes, and their latitudes and longitudes
• Create interactive Folium maps to visualize the venues in districts
• Clean up the DataFrame
• Make calls to the FourSquare API to retrieve the venues in each district
• Try out different distance metrics and algorithms (KDTree vs. BallTre) to determine cluster radii
• Explore the data through visualizations and rankings
• Cluster the data with KMeans to obtain a meaningful represantation of the city

Note: Please view the notebooks through this link to render the Folium maps: https://nbviewer.jupyter.org/