Student Early-Warning System
An explored capstone concept to flag Moroccan school dropout risk using ML, geospatial analysis and an AI assistant.
Voir le code3
data sources
15
brief sections
01
Problème
School dropout in Morocco is hard to act on early because the signals are scattered across World Bank, UCI and local datasets.
02
Approche
Loaded World Bank, UCI and synthetic data into a Bronze layer on DuckDB, planning a dbt + Airflow pipeline feeding clustering, classification and geospatial risk mapping, with a chatbot layer for exploring results.
03
Défi
Scoping a project of this depth against the football-analytics capstone that was ultimately chosen instead.
04
Résultat
A fully specified 15-section brief — parked in favor of Football Atlas, but the architecture pattern carried over directly.
Stack technique
XGBoostRandom ForestK-MeansSHAPFoliumDuckDBdbtAirflowStreamlit