CODEWITHSUNDEEP
javalinear-regression
TheFooBarWay
TheFooBarWay

Reputation: 594

Java Linear Regression

I need to find the best fitting regression line for a set of points. For example for this matrix:

int b [][] = {      { 3, 1, 0, 0, 0, 0, 0, 0, 0 },
                    { 1, 2, 3, 1, 0, 1, 0, 0, 0 },
                    { 0, 1, 2, 1, 0, 0, 0, 0, 0 },
                    { 0, 0, 0, 3, 0, 0, 0, 0, 0 }, 
                    { 0, 0, 0, 0, 0, 0, 0, 0, 0 }, 
                    { 0, 0, 0, 0, 0, 1, 3, 0, 0 }, 
                    { 0, 0, 0, 0, 0, 1, 2, 3, 1 }, 
                    { 0, 0, 0, 0, 0, 1, 1, 1, 2 }, 
                    { 0, 0, 0, 0, 0, 0, 0, 0, 1 }   };

Every number represents the amount of data points (weight I suppose) at that location (where rows are the X axis and Columns are for the Y). I have attempted to use the SimpleRegression class from the apache mathematics library and am having some issues. First, it doesn't appear to support weights. Second I believe that I'm doing something wrong, even for a matrix that is nothing but 1's on the main diagonal the slope/intercept results make no sense.

public static void main(String[] args) {

        double a[][] = new double[9][9];
        for (int i = 0; i < 9; i++)
            a[i][i] = 1;


        SimpleRegression r = new SimpleRegression(true);

        r.addData(a);

        System.out.println("Slope = " + r.getSlope());
        System.out.println("Intercept = " + r.getIntercept());

}

This gives me results that are incorrect. I would assume that this matrix represents the function f(x) = x yet the slope I'm getting is -0.12499..

Could anyone point me at what I'm doing wrong? I have a feeling I'm not only misusing the code but also the mathematics.

Upvotes: 2

Views: 2125

Answers (1)

Chris K
Chris K

Reputation: 1723

As the comments say, addData() expects a 2xN matrix of x y positions or individual x y positions. The following example returns a slope of 1 for a diagonal matrix as expected:

public static void main(String[] args) {
    double a[][] = new double[9][9];
    for (int i = 0; i < 9; i++)
        a[i][i] = 1;

    SimpleRegression r = new SimpleRegression(true);

    addData(r, a);

    System.out.println("Slope = " + r.getSlope());
    System.out.println("Intercept = " + r.getIntercept());
}

public static void addData(SimpleRegression r, double[][] data) {
    for(int x=0; x<data.length; x++) {
        for(int y=0; y<data[0].length; y++) {
            for(int i=0; i<data[x][y]; i++) {
                r.addData(x, y);
            }
        }
    }
}

The example assumes that index 0 corresponds to a position of 0, index 1 corresponds to a position of 1 and so on. If this is not the case you need to add a function to transform index to position.

Upvotes: 1

Related Questions