SJS Sales Intelligence SJS Sales Intelligence

Overview

Developed a predictive sales intelligence system and interactive GIS simulator to model branch performance and analyze competitive attraction in Querétaro.

Highlights

1. 01

   Designed a machine learning pipeline using XGBoost and ElasticNet to predict sales based on spatial competition, urban attractors, and mobility parameters.

2. 02

   Built a high-fidelity React frontend with MapLibre GL JS to display interactive target markets, custom radar charts, and multi-scenario side-by-side comparisons.

3. 03

   Implemented a FastAPI server delivering fast, on-demand inference for new-branch placement simulations with real-time UI updates.

4. 04

   Benchmarked six models (two OLS variants, Elastic Net, Random Forest, XGBoost and SVR) with leave-one-out cross-validation on 82 branches × 19 features and served the selected log-space Elastic Net through an 11-endpoint FastAPI layer.

Stack

• React
• FastAPI
• MapLibre GL JS
• XGBoost
• Zustand
• OSMnx
• H3 Index
• GeoJSON
• Predictive Analytics