GulShair/bike-rental-demand-prediction

🚲 Bike Rental Demand Prediction

Predict daily bike rental demand from weather and seasonal conditions using a Decision Tree Regressor — served through a clean, interactive Streamlit dashboard.

Overview

Urban bike-sharing systems are sensitive to weather and time-of-year. This project builds a Decision Tree Regression model that estimates how many bikes will be rented on a given day based on four inputs: temperature, humidity, wind speed, and season.

The entire workflow — data exploration, model training, evaluation, and live prediction — is available through a multi-page Streamlit application.

Features

• —Dataset Explorer — preview rows, inspect data types, check for missing values and duplicates, view statistical summary
• —Visualizations — histograms, box plots, scatter plots with trend line, and an annotated correlation heatmap
• —Model Training — one-click training with MAE / RMSE / R² metrics, feature importance chart, and actual-vs-predicted scatter plot
• —Live Prediction — number input form with instant demand forecast displayed in a styled card
• —Download Results — export any prediction as a timestamped CSV
• —Assignment Example Loader — pre-fills the required test case (temp=0.5, humidity=0.6, windspeed=0.2, season=3) in one click
• —Clean, responsive UI — sidebar navigation, wide layout, consistent seaborn theme

Tech Stack

Project Structure

bike-rental-demand-prediction/
├── dataset/
│   └── bike_rental_100_rows.csv   # 100-record sample dataset
├── app.py                          # Streamlit UI (6 pages)
├── train_model.py                  # ML pipeline (train, evaluate, predict)
├── utils.py                        # Data utilities (load, clean, info)
├── model.pkl                       # Saved model (generated at runtime, git-ignored)
├── requirements.txt
└── README.md

Installation & Setup

# 1. Clone the repository
git clone https://github.com/your-username/bike-rental-demand-prediction.git
cd bike-rental-demand-prediction

# 2. Create and activate a virtual environment
python -m venv venv
# Windows
venv\Scripts\activate
# macOS / Linux
source venv/bin/activate

# 3. Install dependencies
pip install -r requirements.txt

# 4. (Optional) Train the model from the command line first
python train_model.py

# 5. Launch the Streamlit app
streamlit run app.py

The app will open at http://localhost:8501.

Dataset

Column notes:

• —temp and humidity are normalised floats in [0, 1]
• —windspeed is normalised in [0, 1]
• —season is an integer: 1=Spring, 2=Summer, 3=Fall, 4=Winter
• —count is an integer representing total rentals for that record

This is a sample dataset (100 rows). It is sufficient for demonstrating the workflow but too small for production-grade predictions.

Model

Algorithm: DecisionTreeRegressor (scikit-learn)

Hyperparameters:

The tree is regularised because an unconstrained tree on 100 rows would perfectly overfit the training data and generalise poorly.

Results

Evaluated on 20 held-out test records (20 % of 100).

Custom prediction (temp=0.5, humidity=0.6, windspeed=0.2, season=3): ~206 rentals

Screenshots

Screenshots will be added after deployment.

Future Improvements

• —Collect a significantly larger dataset (thousands of records) for reliable generalisation
• —Compare against Random Forest and XGBoost to quantify the benefit of ensemble methods
• —Hyperparameter tuning via GridSearchCV or RandomizedSearchCV
• —Add time-series features (hour of day, day of week, holiday flag)
• —Cross-validation instead of a single train/test split
• —Deploy to Streamlit Cloud or Hugging Face Spaces

Author

Gul Shair LinkedIn · GitHub