Spark: We use the Iris dataset containing 150 samples and 4 features. With the original dataset split into 80% for training and 20% for test, with a random seed of 13234, how many of the first 20 samples from the test set were correctly classified?