Skip to content

A project that uses AutoGluon to train models to predict bike rental ride counts based on weather forecast.

Notifications You must be signed in to change notification settings

minhtribui153/Bike-Sharing-with-AutoGluon

Repository files navigation

Bike rental prediction based on weather forecast with Autogluon

⚠️ NOTE: It is recommended to run this project on a Conda environment.

This project uses Autogluon to help train all different kinds of models and find which model best suits the purpose for this bike rental prediction.

In this project, Autogluon will use L2 Regression (a.k.a Ridge Regression) to train the models. The models trained by Autogluon will predict Divvy's Bike Rental Dataset based on the weather forecast.

If you would like to generate your own dataset, store the weather dataset by Visual Crossing into the chicago_weatherdata folder and the Divvy Bicycles trip dataset into the divvy_tripdata folder. After that, run the create_dataset.sh command.

The source code for the model prediction in the Bike Rental Predictor notebook file.

Here are the results you will get from this project: Autogluon Models Leaderboard

Prediction Test Data Results

Here are the features you need to take note before testing the model's prediction:

Feature Description Data Type Unit of Measurement
week The week for the date Integer Not Specified
day The day for the date Integer Not Specified
month The month for the date Integer Not Specified
year The year for the date Integer Not Specified
temp The weather temperature Float Degrees Fahrenheit
dew The weather dew point Float Fahrenheit
humidity The weather humidity Float Percentage (0-100)
precip The weather precipitation Float Inches
snowdepth The weather snow depth Float Inches
windgust The weather wind gust Float Miles Per Hour
windspeed The weather wind speed Float Miles Per Hour
sealevelpressure The sea level pressure Float Millibar
cloudcover The cloud coverage Float Percentage (0-100)
visibility The visibility of the area Float Miles Per Hour
solarradiation The radiation of the sun Float Watts Per Square Meter
solarenergy The energy from the sun Float Millijoules Per Square Meter
uvindex The sun's ultraviolet index Float Not Specified

Here are a list of applications that bike rental companies like Divvy can use with this project:

  • Bike distribution (companies can balance the distribution of bikes to areas with better weather)
  • Weather adjustments (if the weather for an area is not suitable for bike renting, companies can use this to close down those areas to save costs)
  • Better Bike Availability (users can benefit from more available bikes, such as during peak hours)

About

A project that uses AutoGluon to train models to predict bike rental ride counts based on weather forecast.

Resources

Stars

Watchers

Forks

Languages