Answer Happy

3. A student compared the results of training a model using the same data and classification algorithm but different training methods, which were: • resub: resubstitution method • hold-out-10%: hold-out method with 90% (training) -10% (testing) split • xVal-10f: 10-fold cross-validation method • LOOCV: leave-one-out cross-validation method The following bar chart shows the average accuracy over 20 trials of using each method, with the error bars representing the standard deviation. 1.00 0.96185 185 185 1852 9414411764705 0.91ERASAARRRRAZ 0.75 Average accuracy over 20 trials 0.25 0.00 resub hold-out-10% xVal-10 LOOCV

(a) Briefly describe each of the above four methods. (4 marks) (b) Compare the four methods in terms of (i) average accuracy, by commenting and explaining the above result (4 marks) (ii) standard deviation, by commenting and explaining the above result (4 marks) (iii) computational cost, based on your knowledge of how each method works. (4 marks) (c) For each of the above four methods, give a scenario in which each method is suitable to be used. Explain your answer. (4 marks)