A dimension reduction model replaces many original variables with fewer composite dimensions that preserve much of the data’s variability. These dimensions become inputs to other models when numerous predictors make meaningful patterns difficult to discern. section 9.1.1
A practical dimension-reduction workflow includes:
- Represent correlated or numerous inputs with a smaller set of components. section 9.1.1
- Use principal component analysis (PCA) as one possible reduction method. section 9.1.1
- Choose a component cutoff near the scree plot’s elbow, where additional components provide diminishing returns. section 9.1.1
- Check whether the resulting components improve interpretability and avoid overfitting without weakening future predictions. section 9.1.1
Purchase prediction
Ten customer-related inputs used to predict time until the next purchase or purchase amount might be represented by three or four components before modeling. section 9.1.1
Table 9.1 Classes of models, their purposes, and examples
|
Class
|
Purpose
|
Example
|
|---|---|---|
|
Regression
|
Predict outcomes represented with continuous data
|
Predicting student test scores
|
|
Classification
|
Predict outcomes represented with categorical data
|
Predict which students will graduate on time
|
|
Clustering
|
Identify underlying patterns and natural “groups” in the data
|
Find clusters of student performance across classes
|
|
Dimension reduction
|
Identify shared variance among predictors to simplify model input
|
Find 2 underlying components to represent scores in 12 classes
|
|
Survival analysis
|
Predict the time until an event occurs
|
Predict the time to withdraw from an intensive graduate curriculum
|
Compare dimension reduction with regression, classification, clustering, and survival analysis by purpose and example. section 9.1.1
