Best random number generator for Learning Vector Quantization (LVQ) first weight

I random first weight for LVQ using java.util.random like this :

//random generator
private double RandomNumberGenerator(){
    java.util.Random rnd = new java.util.Random();
    return rnd.nextDouble();
}

//random data for weight
private void InitializeWeigths(){
    weights = new double[numberofcluster][inputdimension];
    for(int i=0;i<numberofcluster;i++){
        for(int j=0;j<inputdimension;j++){
            weights[i][j] = RandomNumberGenerator();
        }
    }
}

The problem is the result is not good enough, and it depends on the random number for accuracy, can you guys suggest me best random double number generator method or algorithm for this case?