[SYSTEMDS-3150] Outlier Detection via DBSCAN

• Type: Task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: SystemDS 3.0
• Labels: StudentProject
• Epic Link: New Built-in functions
• Assignee: Unassigned
• Reporter: Shafaq Siddiqi
• Created: 04/Oct/21 11:55
• Updated: 17/Mar/22 13:54
• Resolved: 17/Mar/22 13:54

Description

A builtin function to detect the outliers using DBSCAN clustering algorithm (already existing in SystemDS).

The builtin creates a density-based cluster by calling the DBSCAN builtin and returns a Matrix of outlier points.

The builtin could be named as detectOutlier() with parameters like Xtest and borderPoints. The task is to modify the existing implementation of dbscan and return some more parameters such as border points so that these border points could be passed to detectOutlier() function to predict/detect anomalies in the test dataset. 

 

Issue Links

• links to: GitHub Pull Request #1497