Re:Mapping time series to Hilbert curve for finding similar ...
by celoyd on Tuesday May 25, 2010 @01:53AM
(#1376646)
I like sticking stuff on Hilbert curves as much as the next person, but I don’t see how it makes this particular data easier to understand. It looks like you could do the same thing in as little total space with overlapping line graphs, for example. Am I missing something here because I’m not a quant?
[ Parent
| on HN ]
Re:Mapping time series to Hilbert curve for finding similar ...
by abeppu on Tuesday May 25, 2010 @02:13AM
(#1376680)
I don't think the point is to help 'understand' the data better, so much as to help you find similar series, which is admittedly a pretty narrow goal. That said, I posted it because I hadn't seen this tried, and it was creative, not because I think it's the best way to do this. Honestly, I think the most compact way to really approach this problem would be to make a covariance matrix of all the data and display it as a heatmap, which is probably easier too -- but this is also cool, and a bit surprising.
Re:Mapping time series to Hilbert curve for finding similar ...
by celoyd on Tuesday May 25, 2010 @03:09PM
(#1378265)
Oh, I’m not begrudging your posting it; for one thing, it’s got me thinking about Hilbert (and Z-order) curves again. It just seems like an odd application in an area where, I gather, more sophisticated correlation tools are the norm.
Re:Mapping time series to Hilbert curve for finding similar ...
by toppy on Tuesday May 25, 2010 @02:58AM
(#1376727)
Show me your data please.