Prev: Looking for good books
Next: Friedman's TREE(2)?
From: pamela fluente on 18 Jun 2008 13:30
On 27 Mag, 14:52, Patricia Shanahan <p...(a)acm.org> wrote:
> pamela fluente wrote:
>
> ...>   Actually in my case, as you can see from my code I have even been
> > concerned
> >   about *preserving* min and max. Infact I cannot remove that
> > information which is crucial
> >   to the user (for instance in case of prices).
>
> ...
>
> Maybe there is another approach that looks more at the "value" of each
> point for showing extrema and other significant changes in direction of
> the line?
>
> You could assign each point a priority. The priority for recently added
> points would have to be tentative, and adjusted based on subsequent
> data. The priority would depend on things like changes in direction, how
> far the point is above or below the running mean...
>
> The first point, the most recently added point, and extrema over a long
> period are high value. On the other hand, a point that close to the
> straight line between its neighbors is low value.
>
> You could implement this with a 500 element priority queue and a
> circular buffer for the recently added points that need their priorities
> recalculated as new points are added.
>
> Patricia

Thanks Patricia,

that's a nice idea. I will probably implement some of those
criteria
to keep and preserve "key" points.

-P


 | 
Pages: 1
Prev: Looking for good books
Next: Friedman's TREE(2)?