Day 05 — End-to-End Mini Project

A complete machine learning project from raw problem statement to business report. No hand-holding — this is the same sequence you follow on a real engagement.

By the end of today you will have built, evaluated, and documented a churn prediction model that you could walk through in an interview or present to a product manager.

Session Overview

Difficulty: Intermediate–Advanced (all previous topics applied) Reading time: ~1 hour (project brief) | Hands-on: ~4–6 hours Prerequisites: All of Week 01 · All of Week 02 Days 01–04 · A completed EDA from Week 01

Phase	File	Deliverable
Project Brief	01-project-brief	Dataset, project structure, success metric defined
EDA and Cleaning	02-eda-and-cleaning	Clean dataset, written insights, no surprises left
Feature Engineering	03-feature-engineering	Preprocessor pipeline, interaction features, zero leakage
Modeling Pipeline	04-modeling-pipeline	Baseline + 3 models trained, CV scores compared, best model selected
Evaluation and Report	05-evaluation-and-report	Confusion matrix, ROC curve, feature importance, written results section
Submission Checklist	06-submission-checklist	Project passes all quality gates, README complete

Tip

Work through each file in order. Later files depend on objects defined earlier — the dataset is generated in 01-project-brief.md, the pipeline is built in 03-feature-engineering.md, and the final evaluation in 05-evaluation-and-report.md uses the model fitted in 04-modeling-pipeline.md.

Warning

The single most common failure mode in student projects: preprocessing is fitted on the full dataset before the train/test split. Everything in this project uses Pipeline objects to make that mistake structurally impossible. Do not bypass the pipeline.

---

Estimated time: 3–4 hours if you type and run every cell.

Start here: 01-project-brief