I will use confusion matrix to evaluate the decision tree model that I created for this income dataset.

I used confusion matrix to do evaluation because it gives direct comparisons of values including True Positive, False Positive, True Negative and False Negative. By looking at the visualization created by confusion matrix function, we are able to have a quick prediction about the "unseen" data. 

​

I am confident to include that both two classification algorithm, decision tree and pipelined SVC, perform well in the Income dataset based on the test accuracy score, recall, precision and f1-score values. In fact, both these two of classifiers give high score of test accuracy. Notably, the test accuracy score of decision tree classifier and pipelined SVC classifier are the same, which is 0.84. Not only the test accuracy score, the other values of these two classifiers are also surprisingly similar. However, there are also some slight differences about the precision and recall values between these two classifiers. Concretely, when decision tree classifier makes a prediction about if a person make less than $50K a year, it is correct 85% of the time, meanwhile the pipelined SVC classifiers corrects 86% of the time. With only 1% difference in the precision value does not tell anything about whether pipelined SVC classifier works better than decision tree on this Income dataset. In short, both decision tree and pipelined SVC perform well on this Income dataset.

Even though it seems like these two classifiers do the same good job on classifying/prediction on this Income dataset. It does not mean that decision tree and SVC will always perform the same and return the same result since each classifier has its own functions and features.