Provide a scoring module for Watcher

Registered by Antoine Cabot on 2016-02-04

The watcher scoring module is a generic machine learning service. This module can be used to provide for example any predictions or classification that can be used by a strategy.

The module is in charge of storing a "generic learning model" which has been trained and evaluated by an external engine (spark, scikit-learn, TAP, R, ...)

The "generic learning model" is an abstraction that allows the engine to store the results. The aim of the abstraction is to avoid any strong coupling with any external learning solutions. but also to have a better consistency between the different Strategies.

The engine will collect the data from Ceilometer or message broker for online learning, etc.

This module is independent and optional.

Blueprint information

Status:
Complete
Approver:
Antoine Cabot
Priority:
Essential
Drafter:
Jean-Emile DARTOIS
Direction:
Approved
Assignee:
Tomasz Kaczynski
Definition:
Approved
Series goal:
Accepted for newton
Implementation:
Implemented
Milestone target:
milestone icon newton-3
Started by
Tomasz Kaczynski on 2016-03-02
Completed by
Antoine Cabot on 2016-08-29

Related branches

Sprints

Whiteboard

Gerrit topic: https://review.openstack.org/#q,topic:bp/scoring-module,n,z

Addressed by: https://review.openstack.org/289880
    Add specification for Watcher Scoring Module

Addressed by: https://review.openstack.org/289919
    Reformat text to not exceed 79 characters per line limit

Addressed by: https://review.openstack.org/330382
    Add scoring engines to database and API layers

Addressed by: https://review.openstack.org/351577
    Add Scoring Module implementation

Addressed by: https://review.openstack.org/362827
    Add documentation for Scoring Module

(?)

Work Items

Dependency tree

* Blueprints in grey have been implemented.

This blueprint contains Public information 
Everyone can see this information.