80% of iris dataset was used to train and the remaining 20% to test the model. The last part is just an experiment.