Skip to content

Scikit-learn Workflow

Run the Scikit-learn Workflow MicroSim Fullscreen
Edit in the p5.js Editor

About This MicroSim

The flowchart shows the pattern used with every scikit-learn model:

  1. Import the model class
  2. Prepare data: a 2-D X (features) and a 1-D y (target)
  3. Create the model object
  4. Fit: model.fit(X, y) learns the coefficients and stores them in the model
  5. Predict: model.predict(X_new) applies them to new data
  6. Evaluate: model.score(X, y) measures the fit

Below the flowchart is the complete script. The lines of the selected step are highlighted, and a panel explains the step and shows its result for the chapter's eight students (study hours 1 to 8, exam scores 52 to 92). The small plot shows the data, then the fitted line once fit has run, then the prediction.

The results are computed in the MicroSim with the formulas scikit-learn uses, and they agree with scikit-learn 1.8 on the same data: coef_ is 5.7143, intercept_ is 47.0357, and score returns \(R^2 = 0.9970\). For a regression model score is \(R^2\), not accuracy.

The Mistake checkbox changes step 2 to X = df['study_hours']. Single brackets give a 1-D Series, and fit stops with a ValueError. Recent versions of scikit-learn word it as shown in the MicroSim ("Expected a 2-dimensional container but got ... Series instead"). Older versions, and NumPy arrays, give "Expected 2D array, got 1D array instead". Either way the fix is double brackets, df[['study_hours']], or df['study_hours'].values.reshape(-1, 1).

Not included: a switch to other model classes. Only the import line and the create line would change.

How to Use

  1. Press Next to walk through the six steps in order, or click any box of the flowchart or any line of the script.
  2. At each step read the explanation, then the Result panel, which shows what Python holds after that line has run.
  3. At step 5, move the Hours for X_new slider. Watch the predicted score, and read the warning when the hours go outside 1 to 8.
  4. Check Mistake: single brackets on X and step through again to see which line fails and why the later lines never run.
  5. Uncheck the mistake, cover the script, and write the six steps from memory.

Iframe Embed Code

You can add this MicroSim to any web page by adding this to your HTML:

1
2
3
4
<iframe src="https://dmccreary.github.io/data-science-course/sims/scikit-learn-workflow/main.html"
        height="562"
        width="100%"
        scrolling="no"></iframe>

Lesson Plan

Grade Level

High School (11-12) and College Freshman

Duration

10-15 minutes

Prerequisites

The regression equation, the meaning of slope, intercept, and R², and selecting DataFrame columns in pandas.

Bloom's Taxonomy Level

Apply (L3)

Learning Objective

Students will be able to apply the scikit-learn pattern of import, prepare data, create, fit, predict, and evaluate to train a linear regression model and make a prediction.

Activities

  1. Walk through (4 min): Students step through the six steps and write down, for each one, the line of code and what exists afterward: a class, X and y, an untrained model, coefficients, a prediction, a score.
  2. Predict (4 min): At step 5 students predict the score for 6.5 hours by hand from the coefficients and check with the slider. They then find the smallest slider value for which the model predicts more than 100 and explain why that prediction cannot be trusted.
  3. Debug (4 min): Students switch on the mistake, find the step where the script stops, and write the corrected line in two different ways.

Assessment

Given a DataFrame with a column sqft and a column price, students write the lines that train a LinearRegression model and predict the price of a 1,500 square foot house, and explain what model.score(X, y) returns.

References

  1. scikit-learn documentation. LinearRegression.
  2. scikit-learn documentation. Getting Started.
  3. VanderPlas, J. Python Data Science Handbook. O'Reilly. The chapter Introducing Scikit-Learn.