Multi-column encoder with Word2vec

This component uses Python to encode a set of categorical columns into n dimensions.

A Word2Vec model on multiple concatened columns will be estimated, useful to map categories into a continous space.
This procedure is aimed at avoiding the overfitting caused by training a model on variables with and excessive sparsity.
Furthermore, numerical predictors are preferred when working with models which are optimized using gradients.
If desired, it is possible to reduce the dimensionality of embeddings dimensions by applying a PCA algorithm.

Input ports

1. Type: Table
   Training set
   Input columns of the training set
2. Type: Table
   Test set
   Input columns of the test set

Output ports

1. Type: Table
   Training set with dimensions
   Output dimensions (training set)
2. Type: Python
   Word2Vec Model
   Python object
3. Type: Table
   Test set with dimensions
   Output dimensions (test set)