0

I am working an algorithm which I would like to test it on anomaly detection. But unfortunately, it is not yet proper for imbalance dataset, in which instances from one class is much less than the other one.

So the question is

Are there any anomaly dataset which is not imbalance? And how to download it?

Note that, here by anomaly I mean some unwanted item:activity/object,etc, which is , not obvious by looking into a single item. For example, it may be many attacks into a network, but they may not be obvious by looking into any transaction of attacker to the network, but in the overall activity of the attacker.

1

By definition an anomaly is not normal. This means your dataset needs to be somehow imbalanced, since normal defines itself in a bigger context as the regular.

However, You can still take all sorts of anomaly datasets and use different techniques to rebalance them.

Here is a post on some of these techniques: https://www.kdnuggets.com/2017/06/7-techniques-handle-imbalanced-data.html

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.