Back to Projects

DSMarket — Master's CapstoneCase Study

End-to-end data science pipeline for supermarket re-supply strategy optimisation using clustering and time series forecasting.

Data Science
2023
Nuclio Digital School, Barcelona
View live site avnergomes.github.io/dsmarket-dashboard

Project Overview

DSMarket was my capstone project for the MSc Data Science at Nuclio Digital School (Barcelona). The brief: take three years of transactional retail data and design a re-supply strategy that beats the supermarket's existing rule of thumb on a held-out test period.

The pipeline runs EDA, customer and product clustering, demand forecasting per cluster (LSTM and Random Forest baselines), and ships a small dashboard that lets a category manager pick a store + cluster and see the recommended re-supply curve against the actual sell-through.

Development Stages

  • Exploratory Data Analysis on three years of transactional history
  • Customer and product clustering (k-means, hierarchical)
  • Time series forecasting per cluster — LSTM and Random Forest
  • Model deployment and lightweight dashboard for category managers

Tech Stack

Python LSTM Random Forest Clustering Time Series Streamlit

Outcome

Pilot re-supply strategy with full end-to-end pipeline from raw transactional data to actionable per-store recommendations.