"Data science is a doing skill. These hands-on projects — using real datasets from the course — are where it all comes together."
Learning Objectives
By the end of this module, you will be able to:
- Choose a practice project to match your level
- Apply the full workflow to real data
- Know what to focus on for each
How to Use This Module
These are course exercises using real, free datasets (several ship with the source curriculum). For each, run the full loop: ask a question, clean the data, explore and visualize, then communicate what you found. Start simple and level up.
Explain like I'm new: Reading about data science is like reading about swimming — useful, but you don't learn until you get in the water. These projects are the water. Expect to feel stuck sometimes; that struggle is the learning. Every professional data scientist once stared at their first messy CSV wondering where to begin.
A quick way to start any of these: open the dataset, look at the first ten rows, and write down three questions you're curious about. Then pick the one you most want answered and chase it. You don't need a grand plan — curiosity plus the workflow from Module 18 is enough to produce something real. If you get stuck, that's a signal to search, ask, or simplify the question, not a sign you're not cut out for this.
🐝 Honey Production Analysis (Beginner)
Dataset: honey production over time. Practice: loading data, line charts, spotting trends. Goal: How has honey production changed, and where?
Simple example: Chart production per year and per state — a trend and a ranking answered with two charts.
🦠 COVID Case Explorer (Beginner–Intermediate)
Dataset: global COVID time series (in the course data). Practice: time series, cleaning, honest visualization. Goal: How did cases grow over time in different countries?
Common mistake: Comparing raw case counts between big and small countries. Normalize (per capita) for a fair comparison — a classic data-honesty lesson.
🍄 Mushroom Classifier (Intermediate)
Dataset: mushroom features labeled edible/poisonous. Practice: classification, features & labels, evaluation. Goal: Predict whether a mushroom is safe from its features.
🩺 Diabetes Risk Regression (Intermediate)
Dataset: diabetes measurements. Practice: regression, train/test split, metrics. Goal: Predict a disease-progression score from health features.
🚕 Taxi Fare Analysis (Intermediate)
Dataset: taxi trips. Practice: EDA, relationships, feature engineering. Goal: What drives fare — distance, time, both?
Notice how the projects ladder up in skill. The honey and COVID projects build your exploration and visualization muscles (Modules 12–17) on time-ordered data. The mushroom and diabetes projects add machine learning — one a classification (safe vs. poisonous), the other a regression (predict a number). The taxi project sharpens feature engineering: turning raw columns like pickup time and distance into inputs that actually predict fare. Pick one that matches where you are, then climb.
Try this: A good first sequence is honey → COVID → mushroom. Honey teaches you to load data and read a trend. COVID adds cleaning and the honesty of per-capita comparison. Mushroom introduces your first real prediction. By the end of those three, you'll have practiced nearly every skill in this course on data you can actually see and understand.
For each project, aim for the same finished shape: a clear question at the top, a few well-labeled charts, a short honest note on what you found, and — for the ML projects — a train/test split and one evaluation metric (Module 30). That structure is exactly what a real analysis looks like, just smaller. Don't chase a perfect model; chase a complete, honest story that answers your question.
Real-world use case: The COVID project mirrors exactly what public-health analysts did in 2020 — and it teaches a lesson that matters far beyond it. A country with 10,000 cases isn't necessarily worse off than one with 1,000 if it has ten times the population. Learning to normalize per capita here is the same instinct that keeps you honest when comparing sales across regions, clicks across campaigns, or scores across schools.
Practice task: Pick one dataset and take it end-to-end: question → clean → explore → visualize → one-paragraph story. A single finished analysis teaches more than five you started and abandoned.
Data scientist tip: Keep each project in a notebook with clear notes and charts. Polished notebooks become your portfolio — proof to employers that you can actually do the work, not just describe it.
Key Takeaway: Practice with real datasets — honey (trends), COVID (time series, honest comparison), mushrooms (classification), diabetes (regression), taxi (relationships). The projects ladder up from exploration to machine learning to feature engineering. Run the full workflow each time, finish one project completely before the next, and keep polished notebooks as your portfolio.
Further Learning
Datasets and project ideas adapted from Microsoft's Data Science for Beginners (MIT License).