How beltzantoine05/hickathon-6 works
Hi!ckathon 2025 is a PISA mathematics-score prediction project. Its principal workflow cleans and engineers student features, learns a denoised representation, fine-tunes a regression model, and uses AutoGluon ensemble modeling to produce score predictions. The README also describes CatBoost and science/literacy features; these are shown as documented modeling capabilities, though the notebook implementations were not sampled. The current DAE pipeline and preprocessing links are supported by source excerpts; exact wiring into the modeling notebooks is not established.
Main entry points: runner.py DAE pipeline runner · preprocessing.py Cleaning and engineering · autogluon.ipynb AutoGluon ensemble · embedding.py STRATUM embeddings
Components 15
Feature preparation
- Cleaning and engineering
src/hickathon_six/DAE/preprocessing.py - EXO feature setup
src/hickathon_six/dae_exo.py - QUE feature setup
src/hickathon_six/dae_que.py
Representation learning
- DAE pipeline runner
src/hickathon_six/DAE/runner.py - Encoder and DAE
src/hickathon_six/DAE/models.py - Learned representation
src/hickathon_six/DAE/runner.py - Fine-tuned regression
src/hickathon_six/DAE/runner.py
Score prediction
- AutoGluon ensemble
src/hickathon_six/autogluon.ipynb - Math score prediction
src/hickathon_six/autogluon.ipynb
Additional modeling
- CatBoost modeling
src/hickathon_six/catboost.ipynb - STRATUM embeddings
src/hickathon_six/STRATUM-embedding/embedding.py
Other components
- Project user
- Student survey data
- STRATUM glossary
src/hickathon_six/STRATUM-embedding/Glossaire_STRATUM_Clustered_Google.xlsx - Clustered glossary
src/hickathon_six/STRATUM-embedding/embedding.py
Connections 13
- Project user → Student survey data: provides
- Student survey data → Cleaning and engineering: supplies features
- Cleaning and engineering → DAE pipeline runner: prepares inputs
- EXO feature setup → DAE pipeline runner: configures
- QUE feature setup → DAE pipeline runner: configures
- DAE pipeline runner → Encoder and DAE: trains
- DAE pipeline runner → Fine-tuned regression: fine-tunes
- DAE pipeline runner → Learned representation: generates
- Learned representation → AutoGluon ensemble: provides features
- AutoGluon ensemble → Math score prediction: predicts
- Student survey data → CatBoost modeling: provides features
- STRATUM glossary → STRATUM embeddings: provides terms
- STRATUM embeddings → Clustered glossary: exports clusters