Machine learning decision-support system for recommending whether Megaline customers should remain on Smart or migrate to Ultra.
- English documentation: README_EN.md
- Documentación en español: README_ES.md
- Notebook English: notebooks/project_EN.ipynb
- Notebook Español: notebooks/project_ES.ipynb
pip install -r requirements.txt
jupyter notebook notebooks/project_EN.ipynb- Supervised classification for commercial plan recommendation
- Baseline comparison against trivial and interpretable alternatives
- Model selection using validation performance
- Test-set evaluation with accuracy, balanced accuracy, recall, F1, and ROC-AUC
- Bootstrap confidence intervals for performance uncertainty
- Feature importance for business interpretation
The strongest signal for recommending Ultra is mobile data usage (mb_used). The model should be used as a prioritization layer for migration campaigns, not as an isolated decision engine.