Project idea · ungraded

Explain Your House-Price Model

Write an explanation report for the house-price model from the Machine Learning project: which features matter, how each one moves the price, why one particular house got its estimate, and where the explanations disagree.

You will use

Milestones

Build the smallest working version first. Ticks are saved in this browser only.

01Load your trained pipeline and a held-out test set.
02Rank features with permutation importance (mean ± std) and plot the top ten.
03Plot partial dependence and ICE curves for the three most important features.
04Explain five individual predictions with SHAP values and compare them with LIME-style local fits.
05Write a one-page report: findings, surprising effects, and the limits of each method.

Stretch goals

Find counterfactuals: the smallest change that moves a house into a different price band.
Check whether correlated features (rooms and area) split their importance.
Build a small web page that shows the explanation for any house id.

Data

Use the house-price project's data (California housing or Ames). The SHAP library computes Shapley values efficiently for tree models; it is not part of the graded course because it needs compiled dependencies.

Run it yourself

pip install scikit-learn pandas matplotlib shap
python explain.py