/ overview
A full analytical cycle over a raw dataset: profiling, cleaning, imputation and outlier treatment before any modelling took place.
Feature engineering and exploratory analysis surfaced the relationships worth modelling, with visual evidence for each decision.
Several supervised learners were then trained and compared across multiple evaluation metrics rather than accuracy alone, with cross-validation to keep the comparison honest.
/ what it does
- 01Systematic cleaning: missing values, outliers and type normalisation.
- 02Feature engineering driven by exploratory findings, not guesswork.
- 03Multiple supervised models compared under cross-validation.
- 04Evaluation across precision, recall, F1 and ROC-AUC.
- 05Findings written up with reproducible notebooks.
/ numbers
multi-metric
evaluation
cross-fold
validation
end-to-end
workflow
/ stack
Analysis
PandasNumPyMatplotlibSeaborn
Modelling
scikit-learnCross-validation
Workflow
JupyterReproducible notebooks