Store-Expansion Location Scoring
A location-scoring model for retail store expansion on geospatial data; migrated from H2O/Java to Python and settled on PostgreSQL/PostGIS as model engine.
- role
- Python Developer
- period
- 2024 – 2025
Problem
A third-party linear-regression location model in H2O/Java, with a mandate to move it to Python and PostgreSQL+PostGIS and a target of running it on Google Cloud Platform. The model helps experts choose the best location for new stores by predicting key metrics 12 months after opening, including cannibalization of existing stores and competitive effects.
What was built
- Rewrote the model from H2O (Java) to Python; added features and fixed bugs in the vendor code.
- Set the production deployment approach with the client and prepared the environment.
- Analysed GCP’s technical limits for the workload and recommended PostgreSQL/PostGIS instead.
- Used PostgreSQL + PostGIS as the spatial engine for computation and feature engineering, with final variables fed into the regression model.
Outcome
- Reached production on PostgreSQL/PostGIS after the Vertex AI / BigQuery target proved technically unreachable.
- Model available with endpoint documentation under internal Web API.
- Expansion model integrated into systems used in-field by business analytics.