In case I have trained and tested my model with principal components after applying PCA. How should I modify my new instance so that I can use the trained model ?. Let me give an example for a more detailed explanation. Suppose I had 100 features and after applying PCA I decided to use 20 principal components to train a model. Now my new instance will probably have 100 features. So how can I reduce these 100 features (of the new instance) to get 20 principal components for my model to predict?
.
you must project the new data onto the subspace spanned by your principal components ...
https://plot.ly/ipython-notebooks/principal-component-analysis/
.
Burhan Rashid Hussein When you use PCA, it also assigns some weights to all of the hundred variables you have (depending on their influence). So, all you need to do is to save those weights (like a vector) and multiply with any new instance and it should be transformed and then you can run your model on this transformed set of variables.
- Badges
- Science topic