The data scientist’s toolkit, in seven short lessons.

From the shape of a column to a network that learns. Each step asks you to predict first, then shows what real models do: scikit-learn’s trees, forests, k-means and PCA on a 5,000-customer table, and our tools’ own code. About 67 minutes in all, free, with no sign-up.

7 lessons35 steps45 interview questions

  1. 1Distributions and samplingBefore modelling data, know its shape: a bell, flat, skewed or counts.About 9 minutes · 5 steps
  2. 2Confidence intervals and testsAn estimate without its uncertainty is half an answer.About 9 minutes · 5 steps
  3. 3Training, testing and overfittingEvery model is judged on data it hasn’t seen.About 10 minutes · 5 steps
  4. 4Judging a classifierWhen only 7 in 100 customers leave, a model that says “nobody leaves” is 93% accurate and useless.About 10 minutes · 5 steps
  5. 5Trees and ensemblesA decision tree asks one yes-or-no question at a time: easy to read and easy to overfit.About 10 minutes · 5 steps
  6. 6Clustering and PCAUnsupervised learning finds structure with no target to predict: clustering groups similar customers, and PCA squeezes many columns into a few.About 9 minutes · 5 steps
  7. 7How a neural network learnsA neural network is layers of simple units with adjustable weights.About 10 minutes · 5 steps

Start with lesson 1

Built from real models.

  • Predict, then see

    Every step starts with a question to guess. Guessing first, then checking, makes an idea stick far better than reading it.

  • Real models

    Trees, forests, k-means and PCA fitted with scikit-learn on a 5,000-customer table; the statistics, the overfitting curve, the confusion matrix and the network from our own tools’ code. Nothing typed in by hand.

  • Ready for interviews

    Each lesson ends with the interview questions it prepares you for, their answers in our bank, and practice with Prism.