Data Science · Chapter 2 of 43

Data Science Workflow

A typical project loops through: PROBLEM → DATA → CLEAN → EXPLORE → MODEL → EVALUATE → COMMUNICATE → DEPLOY.

Each step feeds back — you often revisit earlier steps as you learn more.

Example 1 (python)

# 1) Define the question clearly
# 2) Collect / query the data
# 3) Clean & wrangle
# 4) EDA (visualise, summarise)
# 5) Model & evaluate
# 6) Present results

The classic loop.

Example 2 (python)

# CRISP-DM is a similar industry framework:
# Business understanding -> Data -> Prep -> Model -> Eval -> Deploy

Same idea, different names.

Key points

Start with a clear question.
Iterate — don't go one-way.
Evaluation belongs in every step.
Communication closes the loop.

💡 Note: Poorly defined problems are the #1 killer of data projects. Always align on the question first.

Data Science Workflow

Key points

📝 Quick Quiz