An eigenvalue measures how important a principal component is by indicating how much variation in the data that component captures. Larger eigenvalues identify stronger components; smaller eigenvalues identify weaker ones. section 9.2.5
Use eigenvalues to interpret PCA results:
- Read principal components in descending strength because PCA orders them by the variance they explain. section 9.2.5
- Treat a component explaining 80% of the original variance as strong and one explaining 5% as weak. section 9.2.5
- Inspect cumulative explained variance to select a component count where additional variance becomes negligible; the illustrated choice is four. section 9.2.5
Choosing components after scaling
After scaling X_train, fit PCA and plot cumulative explained variance. The curve levels off at four components, while PC1 has the highest eigenvalue and the final component the lowest.
