Analysing and predicting inner-city parking space occupancy
summary
City parking data from Dresden VAMOS is scraped into an API and analyzed at 15‑minute resolution over a year; Pandas handles feature engineering (time of day, weekday, holidays, days-to/after holidays) and scikit‑learn models predict occupancy. A DecisionTreeRegressor tuned by grid search yields about R^2 = 0.848 on a train/test split, and the prediction feature importances rank time of day then weekday. The pipeline powers Android and iOS apps and the code and data are on GitHub.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 3:05Scraping parking data
- 7:07Parking patterns over time
- 11:11Features for prediction
- 15:03Testing ML models
- 20:56Prediction vs real data
- 26:45Future predictions and limits
- 30:47Why models differ
- 35:23Open data ethics
- 37:02Open data activism
related talks
43mPython at City Scale
27mSolving Madrid's parking problem with Airflow and Streamlit
20mHow I Solved my NYC Parking Problem with Python
06mD3P: Data-driven Demand Prediction for Fast Expanding Electric Vehicle Sharing Systems
10mHow I solved my NYC parking problem with Python!
21mPredicting Usage for Capital Bikeshare stations based upon Spatial Characteristics
45mMining dockless bikeshare and dockless scootershare trip data
21m"Capacity Prediction" instead of "Capacity Planning":