Partial Least Squares Regression

You can easily download and run the workflow directly in your KNIME installation. We recommend that you use the latest version of the KNIME Analytics Platform for optimal performance.

Here's how the workflow operates:

1. Python Script node generates a dataset with 500 samples, each having 4 feature variables (X_0, X_1, X_2, X_3), and 4 target variables (Y_0, Y_1, Y_2, Y_3). The features are generated based on two latent (hidden) variables "l1" and "l2", which are drawn from a standard normal distribution. The latent variables are hidden factors that influence the observed data but are not directly observable themselves.

2. Then we split the dataset into train and test subsets.

3. PLS Regression is then performed with 2 components and all the feature and target variables selected.

3. A Python view is created showcasing 3 plots.

Top left plot is comparing the real and predicted values for the first component of X and Y. These components are generated from the same latent variable "l1".

Bottom right plot is comparing the real and predicted values for the second component of X and Y.
These components are generated from the same latent variable "l2".

Bottom left plot is comparing the predicted values for the first and second components of Y. These components are generated from different latent variables ("l1" and "l2").

External resources

• Scikit-Learn - PLS Regression
• Scikit-Learn - Compare cross decomposition methods