Data
Constraint-based recommendation system
By Davy AGONMA — Freelance web developer & AI integrator
Problem
A classic recommender simply surfaces the highest scores. In practice a list must also satisfy business constraints: size, diversity, balance across categories, fair exposure for providers.
Solution
A full pipeline built for my master's thesis: user-item matrix, weight matrix through collaborative filtering or latent factors, score reconstruction, then constraint-programming solving with CPMpy to produce feasible lists.
Outcome
A reproducible pipeline from raw matrix to final recommendations, with an explicit and documented constraint model — research work from my IFRI thesis.
Similar projects
Data science & machine learning: what this project covers
The data projects in this portfolio come from my transition into data science: supervised regression, clustering, recommendation systems and constraint programming. They're built in Python, covering the full cycle — exploration, cleaning, modelling, evaluation.
They show what I can do with data today, alongside web development and automation: prepare a dataset, pick a model that fits the problem, and make the result usable inside an interface.
Written by Davy AGONMA, freelance web developer & ai integrator.