Polynomial regression with Apache Maths (Java)

Question

Could anybody help me make a polynomial regression (order 2) with the Apache Math library.

The following data should give this equation: 39.79 x^2 - 497.66 x + 997.45 (computed by Excel with r2 = 0.9998)

// coding style from http://commons.apache.org/proper/commons-math/userguide/fitting.html    

double[] y = { 540.0, 160.0, -140.0, -360.0, -480.0, -560.0, -540.0, -440.0, -260.0, 0.0, 340.0};              
        final WeightedObservedPoints obs = new WeightedObservedPoints();
        for (double figure:y){
            obs.add(1.0, figure);
        }
        final PolynomialCurveFitter fitter = PolynomialCurveFitter.create(2);
        final double[] coeff = fitter.fit(obs.toList());
        System.out.println("coef="+Arrays.toString(coeff));

Here are the regression coefficients provided by the previous code:

coef=[-53.73522460839947, -52.22329678670934, -52.22329678670934]

Obviously, there's something I'm missing...

Thanks for your kind help

Dom

Zielu · Accepted Answer · 2015-02-17 14:41:58Z

8

All your data points are at x = 1.

obs.add(1.0, figure);!!!!

instead of 1.0 there should be x value, if they are evenly spaced from zero than use for loop and ix instead of 1.0.

answered Feb 17, 2015 at 14:41

Zielu

8,6224 gold badges31 silver badges41 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Dominique Over a year ago

Thanks. But according to the doc, the first item is the weight (" ...If no such information exist and all points should be treated the same, it is safe to put 1.0 as the weight for all points"). In addition, what do you call ix ? So I don't see your point.

Zielu Over a year ago

acording to the doc add(x,y) == add(1.0, x,y) if you provide only two arguments the first is x value not the weight. Secondly, you should be worried yourself that you do regression without x values at all. As for ix, I mean it fyou have evenly spread points, the code would look like: for(int ix=0;ix<y.length;ix++) obs.add(ix,y[ix])

Dominique Over a year ago

OK. Got it. You're absolutely right. That gives: coef=[997.45, -497.66, 39.79]. The last one for x^2, the middle one for x and the first one for the constant. Thanks a lot Zielu for your patience. I appreciate.

Collectives™ on Stack Overflow

Polynomial regression with Apache Maths (Java)

1 Answer 1

3 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related