What is the best way to compute trending topics or tags?

Question

Many sites offer some statistics like \"The hottest topics in the last 24h\". For example, Topix.com shows this in its section \"News Trends\". There, you can see the topics

Answer 1

I think they key word you need to notice is "abnormally". In order to determine when something is "abnormal", you have to know what is normal. That is, you're going to need historical data, which you can average to find out the normal rate of a particular query. You may want to exclude abnormal days from the averaging calculation, but again that'll require having enough data already, so that you know which days to exclude.

From there, you'll have to set a threshold (which would require experimentation, I'm sure), and if something goes outside the threshold, say 50% more searches than normal, you can consider it a "trend". Or, if you want to be able to find the "Top X Trendiest" like you mentioned, you just need to order things by how far (percentage-wise) they are away from their normal rate.

For example, let's say that your historical data has told you that Britney Spears usually gets 100,000 searches, and Paris Hilton usually gets 50,000. If you have a day where they both get 10,000 more searches than normal, you should be considering Paris "hotter" than Britney, because her searches increased 20% more than normal, while Britney's were only 10%.

God, I can't believe I just wrote a paragraph comparing "hotness" of Britney Spears and Paris Hilton. What have you done to me?