5. A predictive model (SVM. for example) applied to the validation dataset has classified 90 records as fraudulent (30 correctly so) and 950 has non-fraudulent (920 correctly so) (a) Show the confusion matrix (clearly labeled for 'actual' and 'predicted' cases) [4] ACTUAL FRAUDULENT NON - FRAUDULENT FRAUDULENT 30 60. 30. 920 NON - FRAUDULENT (b) Calculate the error rate and precision (accuracy on the predicted fraudulent cases)? [4] 30920 = 1-0-913. = 0.087~8, 1040 30 30 0.333 33-37 P FP 30+60 90 (c) Assuming the classification threshold used for the above is 0.5. How do you expect the confusion matrix values to change (if at all) when the threshold is increased to 0.7? (4]