Matrix is rank deficient, Multiple Linear Regression Accord.NET

3
votes

I'm working with Multiple Linear Regression in C# using Accord.NET, I followed the example, the method needs 2 arguments inputs which is a 2d array, and outputs which is a 1d array, the two arrays must have the same length.

 public static double[] RegressionLineaire(double[][]input,double[]output)
    {
        double[] coeff = new double[40];
        var ols = new OrdinaryLeastSquares();
        {
            ols.UseIntercept = true;
        };
        Console.WriteLine("inputs length = " + input.Length + " outputs 
        length = " + output.Length);
        MultipleLinearRegression regression = ols.Learn(input, output);
        
        coeff = regression.Weights;

        return coeff;
    }

the inputs and outputs have the same length but I get this exception

System.InvalidOperationException : 'Matrix is rank deficient.'

1
c#linear-regressionaccord.net
Check if your input satisfies conditions HERE. – jsanalytics

1 Answers

4
votes

I just found a solution to the problem, it was not a input shape issue, and i don't know what is exatly the cause of the matrix is rank deficient exception, but i managed de make it work by adding this line.

var ols = new OrdinaryLeastSquares();
    {
        ols.UseIntercept = true;
        ols.IsRobust = true;
    };