Reputation: 7922
PCA on the targets with sklearn?
I am doing multiple target regression, so I want to predict several numbers simultaneously. The numbers are highly correlated, so I think predicting their PC's is a more sensible approach.
Using sklearn's Pipeline is great for daisy-chaining together the MinMaxScaler, PCA, and the regressor I want to use; and then allowing me to call that Pipeline to make predictions.
However, does that Pipeline apply PCA to the inputs only? Is it possible to use the same strategy, but also have it perform PCA on the outputs that I want to predict too?
Upvotes: 2
Views: 322
Answers (1)
Reputation: 9390
No, it is possible to use steps in pipeline only for transforming X argument. You may create additional pipeline for y, where all steps should be transformers, and then feed result of this pipeline as y to a classifier.
Upvotes: 2
Related Questions
- Python PCA sklearn
- Implementing sklearn PCA on limited number of variables in a pipeline
- PCA on data and training with SVM with K-fold CV and Gridsearch
- Scikit-Learn PCA
- Applying PCA to one sample
- PCA analysis considering N-less relevant components
- PCA using sklearn
- Using sklearn Linear Regression and PCA in a single Pipeline
- Using PCA on linear regression
- PCA decomposition with python: features relevances