Ideas for a "suspect changesets classifier"

Posted by naoliv on 17 May 2014 in English (English)

Sometimes we find here in Brazil some imported data from +4 months ago, that nobody saw until now. Usually, these imports are followed by some other changesets deleting the old data + changesets modifying/adjusting the imported data.

We also see some changesets where people purposely/unconsciously delete a lot of data.

Could a Bayesian filter, SVM or something else be used to classify a suspect changeset? Could we use something smart for this task?

Comment from cartinus on 18 May 2014 at 01:42

When using WhoDidIt you can see which changesets contain lots of deletions.

Hide this comment

Comment from naoliv on 18 May 2014 at 02:18

The problem is that I can't manually verify every changeset (and that's why I am wanting some kind of classifier).

Hide this comment

Comment from Nakaner on 18 May 2014 at 13:04

The German user Oli-Wan (a very active German forum member) developes a tool to detect vandalisms and other bad changesets. He has written about his idea/work in German forum. You may contact him in e.g. in German or English.

Hide this comment

Comment from cartinus on 18 May 2014 at 19:02

That is why I mentioned WhoDidIt. Changesets with lots of deletions are specially marked. So you won't have to check them all.

Hide this comment

Leave a comment

Parsed with Markdown

  • Headings

    # Heading
    ## Subheading

  • Unordered list

    * First item
    * Second item

  • Ordered list

    1. First item
    2. Second item

  • Link

  • Image

    ![Alt text](URL)

Login to leave a comment