[SPARK-9904] User guide for ML tree algorithms

• Type: Documentation
• Status: Closed
• Priority: Major
• Resolution: Duplicate
• Component/s: ML
• Target Version/s: 1.5.0
• Assignee: Unassigned
• Reporter: Feynman Liang
• Created: 12/Aug/15 21:21
• Updated: 12/Aug/15 21:51
• Resolved: 12/Aug/15 21:51

Description

The ML Pipeline implementation of trees has diverged sufficiently from MLlib to warrant it's own user guide. Notable features which only exist in ML include:

• Class probabilities (SPARK-6884, SPARK-6885)
• Random Forest Feature importance (SPARK-5133)
• Classification thresholds (SPARK-8069)
• Checkpointing (SPARK-6684)

We should add a new section to ml-guide#algorithm-guides to describe trees in ML pipeline

Issue Links

• is related to: SPARK-6885 Decision trees: predict class probabilities
• is related to: SPARK-6884 Random forest: predict class probabilities
• is related to: SPARK-5133 Feature Importance for Random Forests
• is related to: SPARK-6684 Add checkpointing to GradientBoostedTrees
• is related to: SPARK-8069 Add support for cutoff to RandomForestClassifier
• is related to: SPARK-7772 User guide for spark.ml trees and ensembles
• is required by: SPARK-9668 ML 1.5 QA: Docs: Check for new APIs