3 ms·
This does sound like a neat problem -- and one which I'll be solving soon in my application. Don't want to give away too much, but looks like a basic statistic
by DanielBMarkham 17y ago
This does sound like a neat problem -- and one which I'll be solving soon in my application.
Don't want to give away too much, but looks like a basic statistics text might be a good place to start....
- Perceval 17y agoSo we'd have to set up the null hypothesis that the voting patterns are simply random. What probability distribution are we going to use? Gaussian distribution is probably the wrong one to use, since online voting and online popularity usually conforms better to a power law distribution. So bootstrap a distribution from the data, and then test for correlations? How do we control for the deliberately random behavior of the voting circle sockpuppets?
- karanbhangui 17y agoOn first pass, I thought you were being satirical by just using big terms. Now I see that it's a legitimate question. Carry on.
- rms 17y agoA Bayesian approach probably works much better
- evgen 17y agoThe solution is so readily apparent it does not even require you to know math anymore... Let's reframe the question to see if anyone can figure this one out: "A search engine called Goggle wants to help give you a better view of the web. They use links between web pages to decide which pages will be the top results for a given query; pages with a high authority rank will be features at the top of the search results and will get more traffic from people searching for information. Goggles suspect that spammers have found a way to manipulate the system by creating fake web pages that link to their pages and give them an artificially inflated rank, hence forming 'link farms.' You have been hired to identify suspected farms by looking at link data. Your task may not be as straightforward as it seems however. The caveat is that the spammers may have fashioned their fake pages to look like normal web pages. To create some misdirection, fake users may sometimes remove their links to targeted articles or even link on articles that they have no association with. Sergiy Brin and Larry Page, the creators of Goggles, request that you write a program in Python or C to identify the top five suspected unique link farm rings consisting of at least five pages for each ring." Sound more familiar? It is not quite the same problem, but you will find a lot of good solutions in this well-explored space that also apply to the original problem.