Hacker News Reader

Hi Josh, Very interesting stuff. Can I ask what is the math behind defining the trend? How did you calculate the most trendy words, etc?

I'm working on something similar for a internal tool, so I'm curious to see what your solution is.

Thanks!

View on HN

Of course - I have scripts constantly monitoring the web for potentially interesting keywords. I then check google trends data for that topic/keyword and analyze the gradient of the line of best fit over different timeframes. If the gradient is higher than a certain (arbitrary) amount, then I class it as a trend!

Is the gradient just a linear least squares fit?

Not the author. Seems like it, it doesn't look like Google Trend gives other variables, so just simple 2 var regression I'd guess

Reply on news.ycombinator.com