Spring ‘19 Honda Parking Project: Team 18

Park smarter in Columbus. In collaboration with Honda and the city of Columbus, Ohio.

ParkSmart is a machine learning-based parking guidance system that provides users with a way to find nearby street parking, based on a predictive tool that uses historical data to identify locations where parking is likely to be available. ParkSmart provides turn-by-turn directions to the best parking spots around a user’s inputted destination and helps drivers save time and money, through advanced statistical methods and data science techniques. Using neural network-based time series and forecast models, ParkSmart is designed to predict which zones of Columbus, Ohio, have the highest levels of open parking based on past vehicle trip logs and historical parking meter data. ParkSmart can predict an available spot with 85% accuracy and can help save Columbus drivers an average of 4.5 minutes on each driving trip, nearly 1.2 kilometers of driving at the end of their trips and prevent about 287 grams of carbon dioxide from being polluted into the atmosphere, per trip. ParkSmart was created by a team of Data Science, Electrical Engineering and Computer Science undergraduates at UC Berkeley and developed in collaboration with Honda and the city of Columbus, Ohio.

For more detail, see our:

• Final Paper
• Poster

Quick Start for Web App Demo

Front End Demo is under final-demo/. To run demo locally, cd into final-demo and:

pip install -r requirements.txt

to install all dependencies. (Optionally beforehand, can use virtualenv and pip install in virtualenv. More info on virtualenv from PythonAnywhere.)

Then, to start Flask server:

python demo.py

Then navigate to http://localhost:5000/ (or whichever url is displayed in terminal) to see the app.

If something doesn’t work, please file an issue.

Team

• Clarence Lam
• Derek Topper
• Alex Cho
• Inna Chernomorets