EV Car-Sharing Operations Dashboard

A Streamlit dashboard analysing one year of anonymised events from an urban electric car-sharing service (mobility sector). Each event is either a customer rental or a service operation, where an agent charges a car or moves it off a charging station.

The dashboard answers two questions:

1. How are the vehicles used by customers, and how do agents charge and relocate them?
2. While a car is charging, it may be rented before it is full. Above which battery level is this unlikely enough to send an agent to move it and avoid idle-at-full fees?

Quick start with Docker

docker run --rm -p 8501:8501 yuqixin/assignment-streamlit-app:latest

Then open http://localhost:8501.

Image on DockerHub: https://hub.docker.com/r/yuqixin/assignment-streamlit-app

Run locally

Requires Python 3.12 and uv.

git clone https://github.com/xvnbc/streamlit-app-assignment.git
cd streamlit-app-assignment
uv sync
uv run streamlit run app/streamlit_app.py

Without uv:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements-dev.txt
pip install -e . --no-deps
streamlit run app/streamlit_app.py

The app reads data/rentals.csv by default. Set the DATA_PATH environment variable to use another file.

Project structure

├── app/streamlit_app.py          Streamlit dashboard
├── data/                         Input data
├── src/assignment_streamlit_app/
│   ├── config.py                 Column names, event types, defaults
│   ├── loading.py                CSV reading, column and type normalisation
│   ├── cleaning.py               Deduplication, event labelling, battery imputation
│   ├── filters.py                Date, event type and vehicle filters
│   ├── metrics.py                KPIs, demand patterns, battery distributions
│   └── threshold.py              Rental probability by battery level
├── tests/                        pytest suite
├── Dockerfile
└── .github/workflows/ci.yml      Lint and tests on every push and pull request

Data preparation

cleaning.build_rental_dataset turns the raw file into an analysis-ready table:

• removes duplicated rows, keeping one event per vehicle and start time;
• drops rows with missing values that cannot be reconstructed;
• labels each event as customer_rental, agent_charge or agent_move;
• rebuilds the battery level at the start of agent events from the vehicle's previous event, and the outcome of each charging session from its next event.

Development

uv run ruff check .
uv run pytest --cov

Both checks run in GitHub Actions on every push and pull request.

Dependencies are locked in uv.lock. After changing them, refresh the pip files:

uv export --no-dev --no-hashes --no-emit-project -o requirements.txt
uv export --no-hashes --no-emit-project -o requirements-dev.txt

Build the image yourself:

docker build -t assignment-streamlit-app .
docker run --rm -p 8501:8501 assignment-streamlit-app