Tuesday Tip #26: Get your ML project started quickly 📈

published2 months ago
2 min read

Hi Reader,

In case you haven’t checked out the Data School blog in a while, I’ve published a few new posts:

These are expanded versions of past Tuesday Tips!

🔗 Link of the week

SQL Tutorial for Data Scientists & Data Analysts (free)

Although Python dominated the “Top Programming Languages of 2023”, SQL took first place when ranked by job postings 🥇 (source).

If you’re looking to learn SQL, the tutorial above includes 30+ lessons and 40+ practice problems you can try directly in the browser, some of which were sourced from real Data Science interviews!

👉 Tip #26: Start with a logistic regression model

When faced with a new classification problem, Machine Learning practitioners have a dizzying array of algorithms from which to choose: Naive Bayes, decision trees, Random Forests, XGBoost, neural networks, and many others.

Where should you start? For many practitioners (including myself), the first algorithm to reach for is one of the oldest in the field: Logistic regression.

Here are a few attributes of logistic regression that make it incredibly popular:

  • Runs fast
  • Highly interpretable
  • Doesn’t necessarily require input features to be scaled
  • Can generate a meaningful baseline without any tuning
  • Easy to regularize
  • Outputs well-calibrated predicted probabilities

In other words, it helps you to get going quickly with your Machine Learning project! You can focus your energy on building your initial ML pipeline (from data ingestion to prediction) without spending much computational time or code on model training and tuning.

Understanding logistic regression

Although you can use a ML algorithm without truly understanding it, learning how it works will ultimately help you to develop an intuition for when to use it and how to tune it.

To gain that deeper understanding, I recommend reading this lesson from my Data Science course:

🔗 Logistic regression lesson (Jupyter notebook)

During this lesson, you’ll learn:

  • Why is it called “logistic regression” if it’s used for classification?
  • Why is it considered a linear model?
  • How do you interpret the model coefficients?
  • How does it generate class predictions?
  • What are its advantages and disadvantages?

If you get stuck on any of the concepts in the lesson, the resources listed in my logistic regression guide will help you to get un-stuck!

Tuning logistic regression

If you’ve decided to use logistic regression, you’ll need to tune it in order to maximize its performance. I’ve got a short video that will teach you how to tune logistic regression in scikit-learn:

🔗 Important tuning parameters for LogisticRegression (video)

For more details, check out the scikit-learn documentation.

If you enjoyed this week’s tip, please forward it to a friend! Takes only a few seconds, and it really helps me grow the newsletter! 🚀

See you next Tuesday!

- Kevin

P.S. I thought you said this was a linear system

Did someone awesome forward you this email? Sign up here to receive Data Science tips every week!

Learn Data Science from Data School 📊

Join 25,000+ aspiring Data Scientists and receive Python & Data Science tips every Tuesday!

Read more from Learn Data Science from Data School 📊