Sweet 16: Scooter Traffic Analytics

Where can a scooter go?

Comment

Just all trips on one map
Why trips are in the building? again data quality?
This where people ride scooter!!
Usage of different hours in total and broken down by day of the week
Weekday vs. Weekend Usage, in general weedays > weekends
More longer trips in midnight! - trajectory length vs hour in the day
Wednesday trips
Saturday trips
Trips are centered around
That where scooters are missing!
10 clusters for start points for the entire data
Leisure or utility?

Inspiration

Micromobility is a new global trend empowered with locational technologies.

What it does

Help analyst to draw actionable insights from the automatically collected data.

How we built it

Opensource
Cloud
Data Science

Challenges we ran into

data formats
software bugs

Accomplishments that we're proud of

Data processing pipeline capable of assimilating large data volumes.

What we learned

Getting usable local high-resolution infrastructure data is hard
but you can compensate with advanced analysis for the lack of the data

What's next for Sweet 16: Scooter Traffic Analytics

some sleep and debriefing

Built With

docker
geopandas
jupyter
pandas
postgis
python
qgis

Submitted to

Knoxville City Hackthon 2020
- Winner Data Challenge: Shared Micromobility and Infrastructure

Created by

I did database design, exploration of external data sources, running containers, and making sure that all components work together and results look great.

Alexandre Sorokine
I worked on performing location clustering & destination/path analysis. It was my first time working with GIS systems/data, and I learned so much. This is also my first hackathon, and I had so much fun!

Forrest Shriver
I contributed in data visualization, clustering, data enrichment, prediction, and of course presentations. It was my first time working with GIS systems/data and I learned tons of new stuff. Since I completed my first ever Hackthon, I feel like I CAN DO ANYTHING!

Fan Zhang
Did some analytics on time distribution of rides in pandas. I spent *a lot* of time fighting with QuantumGIS (qgis) to generate the animation seen in our video. Along the way I discovered at 2am that the data was subtly not in chronological order. Also around that time in the wee hours I submitted a bug report for the animation QuantumGIS plugin because the datetime was never rendered as it should have been.

Additionally, spent time feebly looking for good vector data with reasonable metadata, such as structure purpose. Open Street Map data lacks that metadata in Knoxville, and Knoxville's kqgis is closed data (i.e., no API). So we had to abandon doing something clever with route identification based on structure type proximity.

Also, provided some geospatial tool advice along the way -- though not as much as Alex, who was the geospatial MVP.

I also set up some infrastructure, such as our gitlab repo, our Google slide deck, and a Dropbox Paper document for coordinating and documenting activities.

Mark Coletti
Diagonally parked in a parallel universe

Updates

Alexandre Sorokine started this project — Jul 19, 2020 10:03 AM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.