Exercises¶

1. Take the messy dataset in the Practice Corner and write a small function clean_business_df(df) that applies the cleaning steps used above.

2. Modify the worked example to compute total revenue by Region and plot a horizontal bar chart suitable for a slide.

3. (Challenge) Create a small function that detects columns with inconsistent capitalization (e.g., ‘south’ vs ‘South’) and standardizes them.

Hints:
- Use `df.groupby('Region')['Revenue'].sum()` for aggregation
- Use `str.strip().str.title()` to standardize text

Data Loading, Wrangling & Visualization¶

Practical patterns for cleaning messy business data and turning it into persuasive charts. This chapter gives runnable recipes you can reuse in analysis and reporting.

Learning goals

• Load small CSV-like sources into pandas and inspect structure.

• Clean common problems (missing values, wrong types, mis-typed categories).

• Produce a simple summary chart suitable for business storytelling.

Data Loading, Wrangling & Visualisation¶

Preparing Real-World Data for Analysis, Machine Learning, and Business Decisions¶

Learning Roadmap¶

This notebook keeps the full original chapter overview and adds a clearer study path around it.

Welcome to the most honest part of Machine Learning — data wrangling, also known as “90% of the job no one posts about on LinkedIn.” 😅

If math was theory, this chapter is practice with mud. You’ll roll up your sleeves, clean messy data, and make it look like something a CEO would actually want to see in a dashboard.

🧠 Why This Matters¶

Machine Learning models are like gourmet chefs — they can only make good predictions if you give them clean ingredients.

Unfortunately, business data often looks like this:

So before we even think about algorithms, we’ll:

1. Load data from messy sources.

2. Clean it like digital laundry.

3. Transform it into model-ready features.

4. Visualize it like a storytelling pro.

💾 1. The Data Wrangling Trifecta¶

By the end of this chapter, you’ll make data look so clean it could get a job at McKinsey.

📚 Prerequisite: Python Refresher¶

If you’re new to Python or Pandas, don’t panic — it’s easier than assembling IKEA furniture. 👉 Check out my other book: 📘 Programming for Business It covers everything from reading files to basic Python data manipulation.

💡 Tip: You’ll be using libraries like pandas, numpy, and matplotlib. If these look like Pokémon names right now, that book is your Pokédex.

🧩 Practice Corner: “Guess the Data Disaster”¶

Match each messy situation with the tool that saves the day:

✅ Pro tip: Pandas is your Swiss Army Knife for data chaos.

🔍 2. Why Businesses Love Clean Data¶

Messy data → Confused analysts → Wrong dashboards → Angry executives. Clean data → Confident models → Actionable insights → Happy bonuses. 🎉

You’ll soon realize:

Data cleaning is not boring — it’s debugging reality.

For example:

• Missing age? → Estimate with median.

• Wrong gender field? → Normalize text values.

• Negative revenue? → Check for refunds.

• Timestamp errors? → Convert to datetime.

You’re not just fixing numbers — you’re restoring business logic.

🎨 3. Visualisation: Turning Data into Business Art¶

Once you’ve tamed the chaos, it’s time to make your data pretty and persuasive.

This section covers:

• Histograms that show sales trends 📊

• Scatter plots revealing marketing ROI 💸

• Correlation heatmaps for KPIs 🔥

• Dashboards that make execs say “wow” ✨

Remember: “If it’s not visualized, it didn’t happen.” — Every Data Scientist, ever.

💬 4. Business Analogy: The Data Spa¶

Think of your dataset like a customer entering a spa:

When your data leaves this spa, it’s ready for the runway — or your next board meeting.

🧩 Practice Corner: “Wrangle This!”¶

Here’s a messy dataset in Python. Try cleaning it up using what you’ll learn in this chapter:

import pandas as pd

data = {
    'Customer': ['A1', 'A2', 'A3', 'A4', None],
    'Age': [25, None, 300, 40, 32],
    'Revenue': [2000, 500, -100, None, 1500],
    'Gender': ['M', '?', 'F', 'F', 'unknown']
}

df = pd.DataFrame(data)
print("Original Messy Data:")
print(df)

🧽 Challenge:

1. Replace None and ? with proper values

2. Fix negative revenue

3. Correct impossible ages

4. Print the clean version

🧭 5. What’s Coming Up¶

🚀 Summary¶

✅ Data wrangling = preparing the battlefield for ML ✅ Visualization = storytelling for business impact ✅ Clean data = clean insights ✅ Dirty data = bad decisions (and maybe a career pivot)

Remember: “Garbage in → Garbage out” — but in business, garbage often comes with formatting errors.

🔜 Next Stop¶

👉 Head to Data Loading (CSV, Excel, SQL, APIs) to learn how to bring all your data under one roof — without crying over file formats.

Chapter Wrap-Up¶

Your original chapter overview already frames the business reality of messy data well. Keep that mindset while moving through the child notebooks: every technique in this chapter exists to make downstream analysis and machine learning more trustworthy.

Reflection Questions¶

1. Which data-quality issue causes the biggest business risk in your domain?

2. Where in your workflow do data errors usually enter the system?

3. Which visualisation would best persuade a non-technical stakeholder?

8. Interactive Code¶

[120, 145, 138]

['North', 'South', 'West']

9. Guided Practice¶