This project implements a real-time, data-driven pricing engine for urban parking lots using only Python, Pandas, NumPy, and Pathway. The objective is to dynamically adjust parking prices based on demand, competition, and real-time conditions across 14 parking locations.
Model 1 – Baseline Linear Model: Price increases linearly with occupancy.Model 2 – Demand-Based Pricing: Embedded in UDF logic; combines dynamic city features.Model 3 – Competitive Pricing Model: Incorporates geographic proximity and competitor prices to adjust pricing.Dynamic Features: Traffic, queue length, special days, vehicle types.Geospatial Logic: Computes distance between lots.Visualization: Static price trend comparison using Bokeh.
- Python: Core implementation
- Pandas & NumPy: Data manipulation and computation
- Pathway: Real-time logic via
@pw.udf - Bokeh: Interactive pricing visualizations
-
Starts with a base price of $10
-
Adjusts based on:
- Occupancy rate
- Queue length
- Nearby traffic congestion
- Special events or holidays
- Type of incoming vehicle
- Competitor proximity and pricing
Bokeh line plots track pricing across all models for each parking lot. These help visualize how different models respond to demand.
| Model | Description | Implemented |
|---|---|---|
| Model 1 | Linear price w.r.t occupancy | Yes |
| Model 2 | Demand-based pricing | Yes (within UDF logic) |
| Model 3 | Competitive pricing with geospatial logic | Yes |
-
Data Ingestion: 73-day simulation over 14 lots with time-stamped intervals.
-
Feature Extraction: Parse occupancy, traffic, vehicle types, etc.
-
Pricing Logic:
- Model 1: Linear Occupancy
- Model 2: Demand Score via city & vehicle factors
- Model 3: Adjust with competitor prices and distance
-
Real-Time Processing:
- Pathway UDF for model execution
-
Visualization:
- Bokeh pricing plots (SEE NOTEBOOK)
This project is part of Summer Analytics 2025, hosted by Consulting & Analytics Club × Pathway.
HIMANGSHU SHEKHAR