Tracking and analyzing bird scooter locations across Tel Aviv

• Lambda function for sampling bird locations using the bird.co internal API.
• Heatmap visualization of bird locations during a given day.
• Analyze bird usage and revenues per day.
• Predict the chances of finding a bird at a given time and location.

One day of TLV Birds location heatmap - demo

Full timeline GeoJson - link

Key concepts

• Data is collected from publicly available sources. The birds are tracked using the bird/nearby API, which is basically the same as refreshing the app and registering locations every minute.
• The collected data is somewhat noisy. Inconsistent GPS signals and maintenance operations might look like bird rides. The data is cleaned by using a few heuristics:
  • Minimal bird ride is at least 100 meters long.
  • No ride is shorter than 1 minute.
  • Rides consume at least 1% of battery power.

Usefull commands

Copy data from s3 - aws s3 cp s3://birds-locations/${date} ./data/${date} --recursive

Generate heatmap data - node process.js ./data/${date} > heatmap_demo/heatmap_data.js