alglib, C#, VS2010 and that Excel curve....

ReinhardtB · **Joined:** Wed Sep 08, 2010 9:05 am **Posts:** 3

Hi all -

I'm quite obviously a n00b (1 post! ;) ) so please be gentle! I'm also don't resemble anything close to a mathematician, and the users want bloody trend lines!!!! :roll:

Using the above tools, would it be possible to generate a best-fit (NOT necessarily dissecting all points), smooth trend curve on a chart?? I've looked at lsfitlinear which I understand should be the answer, but I can't for the life of me figure out what the usage is in c#?!?

Your input would be GREATLY appreciated, and free beer doled out if you ever find yourself in South Africa! :mrgreen:

Cheers,
Reinhardt

Sergey.Bochkanov · **Joined:** Fri May 07, 2010 7:06 am **Posts:** 927

Yes, you can use lsfit unit, but it requires you to specify some function which will be fitted.

If you want to have flexible fit, I recommend you to use spline fitting (see http://www.alglib.net/interpolation/spline3.php). You can use Spline1DFitCubic function which accepts only one "tweak" parameter - number of degrees of freedom. You can also use Spline1DFitHermite, it is more flexible, but it have one additional requirement - number of degrees of freedom must be even.

See http://www.alglib.net/translator/man/ma ... line1d_fit for examples on this subject.

If you want to use linear/polynomial fitting, take a look at the BarycentricFit... functions.

ReinhardtB · **Joined:** Wed Sep 08, 2010 9:05 am **Posts:** 3

Fantastic - thanks for the quick reply!

I have my code as follows:

Code:

            double[] y = dataPoints.Select(p => (double)p.Result).ToArray();
            double[] x = new double[dataPoints.Count];

            for (int i = 0; i < x.Length; i++) // it is always a counter...right?
                x[i] = i;

            int resultIndicator = -1;
            alglib.spline1d.spline1dinterpolant spline = new alglib.spline1d.spline1dinterpolant();
            alglib.spline1d.spline1dfitreport report = new alglib.spline1d.spline1dfitreport();

            alglib.spline1d.spline1dfitcubic(ref x, ref y, dataPoints.Count, 4, ref resultIndicator, ref spline, ref report);

            if (resultIndicator > 0) // success
                for (int i = 0; i < dataPoints.Count; i++ )
                    dataPoints[i].TrendPosition = (decimal)alglib.spline1d.spline1dcalc(ref spline, (double)i);

does this look right?

Sergey.Bochkanov · **Joined:** Fri May 07, 2010 7:06 am **Posts:** 927

Almost right. I recommend you to:
* pay attention to number of degrees of freedom. 4 may be too small for highly nonlinear trends.
* calculate spline values at even intervals spanning [min x, max x], not at the points themselves.

However, these are mostly cosmetic changes. Your code looks right.

ReinhardtB · **Joined:** Wed Sep 08, 2010 9:05 am **Posts:** 3

Thanks Sergey - it's working!

I can not set the degrees of freedom any lower than 4 (I suppose that's by design), but higher than 4 makes a significant difference to the curve! I'll have to add that in as a parameter.

Thank you so much - it works spectacularly well!

I hope Google finds this thread, I struggled my ass of to find an answer to this!

Thanks again!

forum.alglib.net

Forum rules

alglib, C#, VS2010 and that Excel curve....

Who is online