Ride Sharing Demand Competition - Kaggle

Overview

This repository contains the code and resources for participating in the Ride Sharing Demand Competition on Kaggle. The competition involves predicting the demand for ride-sharing services based on various features such as time, weather conditions, and other relevant factors.

Data

The dataset provided for the competition includes historical ride-sharing data with features such as:

• Datetime
• Season
• Holiday
• Workingday
• Weather
• Temperature
• Feeling Temperature
• Humidity
• Windspeed

Additionally, the dataset contains the target variable 'count', representing the number of ride-sharing bookings at a given time.

File Descriptions

• train.csv: Training dataset containing historical ride-sharing data.
• test.csv: Test dataset for making predictions.
• sampleSubmission.csv: Template file for submitting predictions to Kaggle.

Code

The code for this project is organized into Jupyter Notebooks:

• project-template.ipynb: Notebook containing exploratory data analysis and feature creation. Notebook for training machine learning models using AutoGluon's Tabular Prediction. Notebook for post-processing predictions to ensure validity.
• project-template.html: HTML file of the above notebook

Results

The model trained using AutoGluon achieved competitive performance on the Kaggle leaderboard, with a root mean squared error (RMSE) metric indicating the accuracy of predictions.

Dependencies

The following Python libraries are required to run the code:

• pandas
• numpy
• matplotlib
• seaborn
• AutoGluon

Acknowledgements

• The dataset for this competition is provided by Kaggle.
• AutoGluon library was used for training machine learning models.