How mathblender17/amazon-ml-challenge-2025-price-prediction works
This repository is a notebook-based solution to the Amazon ML Challenge 2025 product-price prediction task. Its principal workflow transforms product listing text and, in earlier experiments, images into features, trains and evaluates price models, and produces predictions. The README documents three iterations and the final text-focused ensemble; source excerpts were unavailable, so the graph reflects documented capabilities rather than unverified notebook-level call wiring.
Main entry points: Version 3 optuna optimised amzonmlchall.ipynb Price regressors · Version_O_amazon-ml-challenge-2025.ipynb Feature fusion · Version_1_strat_imd_down_amazon-ml-challenge-2025.ipynb Text cleaning · submission_lgbm_tuned.csv Tuned submission
Components 12
Text Features
- Text cleaning
Version_1_strat_imd_down_amazon-ml-challenge-2025.ipynb - Text embeddings
Version 3 optuna optimised amzonmlchall.ipynb - Text alternatives
Version_O_amazon-ml-challenge-2025.ipynb
Image Features
- Image preparation
Version_1_strat_imd_down_amazon-ml-challenge-2025.ipynb - Image embeddings
Version_O_amazon-ml-challenge-2025.ipynb
Modeling
- Feature fusion
Version_O_amazon-ml-challenge-2025.ipynb - Price regressors
Version 3 optuna optimised amzonmlchall.ipynb - Optuna tuning
Version 3 optuna optimised amzonmlchall.ipynb
Evaluation and Output
- Model evaluation
Version 3 optuna optimised amzonmlchall.ipynb - Price predictions
Version 3 optuna optimised amzonmlchall.ipynb - Tuned submission
submission_lgbm_tuned.csv
Other components
- Product listings
Connections 13
- Product listings → Text cleaning: provides text
- Text cleaning → Text embeddings: feeds features
- Product listings → Image preparation: provides images
- Image preparation → Image embeddings: prepares images
- Text embeddings → Feature fusion: supplies text features
- Image embeddings → Feature fusion: supplies image features
- Feature fusion → Price regressors: feeds models
- Text embeddings → Price regressors: feeds text models
- Price regressors → Optuna tuning: is tuned
- Price regressors → Model evaluation: is evaluated
- Optuna tuning → Price predictions: selects model
- Price predictions → Tuned submission: writes predictions
- Product listings → Text alternatives: provides text