The classic binary classification, the Kaggle's 'hello-world' (https://www.kaggle.com/c/titanic).
You can find my step-by-step notes here. The project is split into two standalone jupyter-notebooks.
| desc | link | notable |
|---|---|---|
| Examine the data. Build a robust and reproducible preprocessing pipe with feature transformation and selection. Visualize the data. | https://github.com/olszewskip/Titanic/blob/master/preprocess_visualize.ipynb | custom-Transformer-classes, ColumnTransformers, FeatureUnion, Pipeline; feature-importance-tests; PCA, TSNE |
| Fit and tune classifiers using grid-search with cross-validation. Report accuracy, confusion-matrix, ROC-curve. Select best based on independent validation score. | https://github.com/olszewskip/Titanic/blob/master/classify.ipynb | GridSearchCV, LogisticRegression, NaiveBayes, GradientBoosting, AdaBoosting, VotingClassifier |
plus a snippet which prepares Kaggle-submittable csv: https://github.com/olszewskip/Titanic/blob/master/prepare_submission.ipynb