[SPARK-9540] Optimize PrefixSpan implementation

• Type: Improvement
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.5.0
• Fix Version/s: 1.5.0
• Component/s: MLlib
• Target Version/s: 1.5.0
• Assignee: Xiangrui Meng
• Reporter: Xiangrui Meng
• Created: 02/Aug/15 16:16
• Updated: 05/Aug/15 05:29
• Resolved: 05/Aug/15 05:29

Description

Current `PrefixSpan` implementation contains some major issues:

1. We should expand the prefix by one item at a time instead of by one itemset.
2. Some set operations should be changed to array operations, which should be more efficient.

Issue Links

• relates to: SPARK-6487 Add sequential pattern mining algorithm PrefixSpan to Spark MLlib
• links to: [Github] Pull Request #7937 (mengxr)