程序代写 COMP90049 Introduction to Machine Learning (Semester 1, 2022) Workshop: Wee

School of Computing and Information Systems The University of Melbourne
COMP90049 Introduction to Machine Learning (Semester 1, 2022) Workshop: Week 11
1. For the following dataset:
ID Outl Temp Humi Wind PLAY

Copyright By PowCoder代写 加微信 powcoder

TRAINING INSTANCES AshhFN BshhTN CohhFY DrmhFY ErcnFY FrcnTN
TEST INSTANCES GocnT?
Classify the test instances using the ID3 Decision Tree method: a) Using the Information Gain as a splitting criterion
b) Using the Gain Ratio as a splitting criterion
2. Let¡¯srevisitthelogicbehindthevotingmethodofclassifiercombination(usedinBagging, Random Forests, and Boosting to some extent). We are assuming that the errors between the all classifiers are uncorrelated
(a) First,let¡¯sassumeourthreeindependentclassifiersallhavetheerrorrateofe=0.4, calculated over 1000 instances with binary labels (500 A and 500 B).
(i) Build the confusion matrices for these classifiers, based on the assumptions above.
(ii) Using that the majority voting, what the expected error rate of the voting ensemble?
(b) Now consider three classifiers, first with e1 = 0.1, the second and third with e2= e3= 0.2.
(i) Build the confusion matrices.
(ii) Using the majority voting, what the expected error rate of the voting
(iii) What if we relax our assumption of independent errors? In other words, what
will happen if the errors between the systems were very highly correlated instead? (Systems make similar mistakes.)

程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com