Skip to content

Navigation: Part Index | Main Index | Finding Your Idea -->

Why a Project?

Requires: What Is Artificial Intelligence? · CRISP-DM

Motivation: You have just seen what data science is and how CRISP-DM structures a project end to end. But there is a gap between understanding a process in the abstract and actually running it from start to finish. The most reliable way to close that gap is to work on a problem you genuinely care about.

You'll see why project-based learning builds transferable skills, what makes a project valid in this course, and why the process matters more than the specific solution you end up with.

Table of Contents

Learning Sticks When the Problem Is Yours

When was the last time you learned something well enough to use it on your own? Chances are you were solving a problem you cared about.

Skill learning transfers when you care about the skills.

  • When you encounter the same technique in a new situation, you already know it can do something useful.
  • Motivation matters too: when the problem is yours, you push through the challenging parts.

A project you genuinely care about gives you this, whether you work alone or in a group. You apply the process from this course (CRISP-DM) to a question from your own field, using data that actually matters to you. The result is something concrete you can point to, explain, and build on.

Working on a project you care about is an invitation. Students who take it up tend to get more out of the course and leave with something to show for it.

What Makes a Project Valid

In practice, a good data scientist does not start by asking "which model should I use?" They start by asking "what problem are we solving?".That order matters.

In real work the problem determines what the solution should look like, not the other way around.

Although this course teaches them, machine learning models are just one possibility among several.

For your project, that means: the goal is to learn the process, not to produce a specific type of output.

Three conditions make a project valid:

  1. There is a real problem. Something you could state in one sentence: a decision that takes too long, a quantity you estimate by hand, a pattern you suspect exists but have never verified. The problem does not have to be large or ambitious.

  2. There is data you can access. Tabular measurements, sensor logs, images, a spreadsheet you maintain yourself: All good. The data does not have to be large or clean at the start.

  3. You work through the phases and can justify your decisions. This is the core requirement. For each phase of CRISP-DM (see 🖝 CRISP-DM), from business understanding through evaluation, you document what you did and why. That is what you will be asked to explain and defend.

What this means in practice: a project that ends with a rule-based threshold, a well-reasoned automation, or a simple formula counts if you applied the process and can justify why that solution fits the problem better than alternatives. The requirement is the same for ML-based solutions: If they are unnecessarily complex for the problem, that's actually not good.

A question you can state in one sentence, and data you can access: That is all you need to start. The process will guide the rest.

Don't wait until you feel confident. Start.

Summary

  • Skills transfer better when learned on a problem you care about: a project creates that effect, whether you work alone or in a group.
  • The project is about learning the CRISP-DM process, not about producing a particular type of solution.
  • A valid project has a real problem, accessible data, and documented decisions at each phase.
  • Machine learning is one possible outcome. A rule-based solution or simple automation also counts if the process was applied and decisions are justified.

As always: Happy learning, happy life! 🫶

Navigation: Part Index | Main Index | Finding Your Idea -->

Script v1.4 (2026-06-10) · FGN