Reference no: EM132511621
Machine Learning Assignment -
Topic includes Data Preprocessing, Random Forest, Neural Network. Questions has to be completed by solving them manually i.e., derive and numerical computation etc.
Question 1 - A bank has decided to use machine learning to determine if a person should be granted a loan. Assume that the database of the bank contains the following data, where the last column shows the credit history of a potential client.
|
Age
|
Married
|
Unemployed
|
Credit
|
|
39
|
yes
|
yes
|
bad
|
|
25
|
no
|
yes
|
bad
|
|
35
|
yes
|
no
|
good
|
|
19
|
yes
|
yes
|
bad
|
|
25
|
yes
|
no
|
good
|
|
20
|
no
|
no
|
bad
|
|
34
|
no
|
no
|
good
|
|
26
|
No
|
no
|
good
|
|
22
|
No
|
no
|
bad
|
|
18
|
yes
|
no
|
good
|
(a) Assume that the maximum height of a tree is two, that is a tree may only consist of the root and its two children. Use Random Forest with mtry = 1 to construct a forest of 3 trees for the credit dataset above. You are not supposed to use any ready made implementation of Random Forest.
(b) How is the credit worthiness of the following new customers classified by your forest? Clearly explain why you obtain a given classification.
|
Age
|
Married
|
Unemployed
|
Credit
|
|
28
|
yes
|
yes
|
?
|
|
23
|
yes
|
no
|
?
|
|
27
|
no
|
no
|
?
|
Question 2 - The following data set describes whether a given coordinate is on your own property or not.
|
(x, y)
|
Class
|
|
(10.5, 5)
|
unowned
|
|
(7.5, 6)
|
unowned
|
|
(2, 5)
|
owned
|
|
(3, 4)
|
unowned
|
|
(3.5, 8)
|
owned
|
|
(6.5, 6.5)
|
owned
|
|
(1.5, 6)
|
owned
|
|
(6.5, 8)
|
owned
|
|
(8, 4)
|
unowned
|
|
(10.5, 9)
|
owned
|
(a) Construct a neural net with only one tanh node that perfectly differentiates between owned and unowned land. Describe how you find the weights in the neural net.
(b) When looking at a new map, it was found that also the points (5, 9) and (5, 10) are unowned. Construct a new neural net that gives perfect classification for both the data above and these new points.