Pergunta

Is there any software that enables me to perform LOF outlier detection towards a 2D dataset? I just implemented LOF from the original paper and want to check whether my results are correct. So far I couldn't find any tool or online service.

Foi útil?

Solução

You should check rapidminer application. And add anomaly detection plugin.

Outras dicas

Did you check the Wikipedia article on Local Outlier Factor? It mentions ELKI, which has a LOF implementation (with index support, so O(n log n) for large data sets).

Note that many people overlook the reachability-distance in LOF, and only approximate it. The ELKI implementation supposedly is correct, so it can help you well for testing your results.

Rapid Miner considers the IP address attribute of DARPA 1998 dataset as polynomial type data ! But LOF is not optimized for that. Interestingly RAPID MINER cannot handle such huge network traffic datasets. Any other implementation available ?

Licenciado em: CC-BY-SA com atribuição
Não afiliado a StackOverflow
scroll top