Anomaly Detection Engine

From OpenStreetMap Wiki
Jump to: navigation, search

Anomaly Detection Engine was one of the Google Summer of Code/2012/AcceptedProjects, a project with the aim to create software that is capable of identifying "bad" map edits: vandalism and accidental changes alike. With the help of machine learning techniques the software should catch these edits and provide a simple web interface for reviewing the edits and to take action.


Weeks 1-2
(May 21 – Jun 3)
Bonding with the community, creating specification of desired results and methods of training.
Week 3
(Jun 4 – Jun 10)
Creating design documentation including libraries I'll be planning to use and outline of basic code structure.
Weeks 4-5
(Jun 11 – Jun 24)
Implementation of the chosen machine learning algorithm.
Weeks 6-7
(Jun 25 – Jul 8)
Customizing libraries for parsing map features and elements if necessary; transforming data from changesets into values that can be used as input for the learning algorithm. At this point the learning algorithm is capable of classifying changes, albeit it's not producing useful output yet. (Goal for Mid-term Evaluation Period)
Week 8
(Jul 9 – Jul 15)
Writing tests for the learning algorithm and creating the web interface.
Weeks 9-10
(Jul 16 – Jul 29)
Training of the learning algorithm.
Weeks 11-12
(Jul 30 – Aug 13)
Testing and finishing documentation.

Project resources


Mentored by Derick Rethans (OSM wiki user page)