[POC] Framework for auto-encoder based approaches for recommendations for NPM

[sara-02]

User Story:

As an consumer of Fabric8-Analytics stack report, I should be able to get companion/outlier insights for my NPM stack via the new approach.

Acceptance Criteria:

Prepare the framework for auto-encoder based approaches as described in this issue.

Task List:

• Data Cleaning
• Data Preparation: Generate the Rating Matrix and the Context Matrix
• Feature engineering and selection
• Create the Supervised Auto-encoder Framework using Keras(not required since the CVAE approach worked).
• Hyper-parameter tuning
• Accuracy evaluation

Note: The details about the approach can be found in this issue, and in this doc

[rootAvish]

The CVAE approach worked, and has its results documented here: https://docs.google.com/spreadsheets/d/1OKq1BvPHUzKwYrL8XFWZAgkKC7jcaW0QyY0vleq0wV0/edit?usp=sharing, we did not go forward with implementing the supervised autoencoder approach as it would perform inferior to this approach and was a fallback in case the other did not work. Work around creating a post-filter business logic for the same will be carried out in this sprint.