~/projects

Dynamic Pricing Model for Parking

A machine learning pricing system that optimises parking lot revenue under fluctuating demand, with A/B testing validation and an interactive Streamlit dashboard.

PythonXGBoostStreamlitPandasScikit-learn
Source
Dynamic Pricing Model for Parking

Overview

An end-to-end dynamic pricing system modelling demand response and revenue optimisation for parking availability.

What it does

  • XGBoost regression on time-series and behavioural features (hour-of-day, lag occupancy, rolling averages), MAE of 0.13
  • Price elasticity simulation with A/B testing framework comparing static vs dynamic strategies
  • Sensitivity analysis across elasticity and demand regimes
  • Interactive Streamlit dashboard for real-time forecasts and pricing recommendations
  • Identified strategies increasing simulated revenue by ~18%

The lesson

The A/B framework revealed that elasticity assumptions dominate model performance more than feature engineering. Getting the demand curve right matters more than squeezing the last bit of XGBoost accuracy.