[SPOT-196] [ML] Spot LDA Wrapper refactoring

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Labels: spot-ml
• Epic Link: [ML] Spot-ml code refactoring for better modularity
• Assignee: Ricardo Barona
• Reporter: Ricardo Barona
• Created: 10/Jul/17 19:10
• Updated: 11/Oct/17 17:31
• Resolved: 11/Oct/17 17:31

Description

Modify SpotLDAWrapper object to decouple functionality.

• Create new object SpotLDAModel as a wrapper for Spark DistributedModel and LocalModel that can Save or Predict (only LocalModel).
• Create new object SpotLDAHelper that can format LDA corpus, topicDistribution and topicMix.
• Create new object SpotLDAResults that contains 2 attributes, one is the document distribution over topics and another one for word distribution over topics required for spot-ml scoring.
• Modify SpotLDAWrapper to be able to run or load, when run method is executed, run LDA and return a new SpotLDAModel. When load method is executed, load an existing model.

Issue Links

• links to: GitHub Pull Request #104