In supervised machine learning, when calculating prediction on a dataset in r split in train and test I get the same prediction using logistic regression, support vector machines and k-nearest neighbour. Is it a good thing or have I done something wrong?