Nirma University

Institute of Technolog;
Semester End Examination (lR),
December _ 20lg
Master of Computer Application,
Semester _ V
3CA1566 Machine Learning and
Applications

Roll/
Supervisor's initial
Exam No.
with date
Time: 3 Hours
It4a<. Marks: 100
Instructions: 1. Attempt all questions.
2. Figures to right indicate full marks.
3. Use section-wise separate answer
book.
4. Draw neat sketches wherever nu."riur.
5. Assume suitabre data whererur n"*Jry
and mention the same.
6. Use of non-programmabre catcutatoiittoiveo
as pu,. ou,. uiiilrsity rutes.
7. Graph paper required.

Section I
Q-1. Answer the following questions: -
A' Murtipre choice euestions. Select the
correct choice(s) and grve [161
CO3BL2 justification for the answer. the (04)

(i) which of the folowing opti.ns are regression

(a) predict the pollutironlevel tasks?
for
(b) predict the percentage fo,
a " "1?y
(c) predict the GDp for a Nation "tua"nt.
ror trre next year.
(d) predict the sales profit of
a company ro, tlr* rrohday season.
(ii) which of the following are nomi'al
features
(a) gender of a person (b) pr:ice of a product
- (c) cognitive revel of a student '1dy6esign"til, or1r, employee.
B' Evaluate the following two confusion matrix
co2iil,s two ciassification moJels tables given below for of (12)
ao trr.-ffi*irrg.
""a
Model-l confusiorEaF Model- ?;!,''*Xtlt"*
I Classified as
ActuaI AS
A (+ve) B (-ve)Actual A (+ve)
Class B (-ve)
Class
A (+ve) 60 05 A (+ve)
B (-ve) 50 15
55 55 B (-ve) 40 70
Note: (+ve mean"
m posEffiGd _ve meals negative
ctass;---
(i) Calculate the classification accuracy,
sensitivity and specificity for
both the models,
- Draw the bar
(ii)
.chart showing the sensitivity and specificity and
accuracy for both the models.
(iii) Discuss which moder is showing
better performance in terms of
sensitivity and specificity. Write your observations.

P.'ir .l- s[ ",

 'rCA i {.*66-Me,cha.,.t Le'ivr"i-6
and Appr*'e..fr 6-r1s
Q-2. the following
A. {19wer questions:
Differentiate between Find_S
CO2BL4 with a suitable example. and Candidate Elimination
Algorithm [:i
A.
CO2BL4 Yhll is Neurar Networksa co-l?Iland,contrast.the various
Networks types (06)
:1"ffi1J." "."r,i-,1"il.", with "rrit.ur. diasrams -and
B. The neuron model given
CO3BL3 below takes two binary
X2. rhe values of xf yflr.-d inputs X1, (06)
iz".'iol'.rf
""arun"tloi-,",'rrf"t"";,h;;;rra!i9n funciion: h1x;
is the binary threshold
=l;;;'.rrd h(x) = o if x>=0.
The corresponding rr.rgfri",'"WO:rO,..yr:.;il
specified on the respectiie
inputli.,".. which otirre ..ra W2=+2O are
folrowing logical
31Ti",Ht"3t;"Hd "; N-oR,"i""* ,r," ".;;';mputea Show- the

C. Discuss the working

CO 1BL2 of Gradient Descent algorithm
regression with multipte for
linear (06)
"aria\p ,"r."g'" .ff#or" example.
C.
CO 1BL2 ilffH i,t:"-Jffi!I'J,##il;l'!Slrj method in a regression (06)

Q-s. Answer the following questions:

CO3BL3
frr'Ljfl:1, :J',$3;"d*i1;"*fu
3r.":Hs;[#,['r'
-
*l am oun t or arar5. received
j*.:-,,-"1,.i;:-^':l.trf+'Jfi 1ff';ffi:*(i;i
s
I Sl I

tl;t";1,:I"fJi;^;"i";ii:;:in,ri+3ili!d':':il';"HT#
$m;i.%::.,::::LT,t{:*T1q-."#,i:""""""',f,
Variable .rv*vwrrrEi Laore. X ts the."iXA#S:l:
(predictor) and y is
i dependent
i
regressio n equahon is -the- variable. The linear
Y=00+$rX.Dorh
ProgressivE- SaIary per annum
Performance received through campus
Index (ppl) ptacement in lacs (Sil)
(x)
7.5
5 i
9 ,i
L2 E

6.5 a
a
I
e

i
E 8
5.5
10
i

_1!_
6.9
9.2 1s_
q
a) Determine a linear regression
data. Find the value for the model equation to represent
two parameter" go '"t *J
this
r"r"

flnfie Z i,,1 :::.

 3(Ri lfr;g..- 'lrtr,rc-X",;
ty: Le._-r{l1,yg .\^r\ 4 pp.L,cxh irr,,.,g
b) Plot new equation.
c) If a student has 6 PpI then based upon this study, what can be
the predicted salary that he may receive?
d) Draw the box plot for the values of the attributes given in the
above mentioned table.
Section - II
Q-4. Answer the following questions:
A. What is cluster analysis? Discuss it briefly. [161
(0s)
CO 1BL2
B. Discuss the difference between fo[owing two hold-out methods: 10- (0s)
CO2BL5 fold cross validation and leave-one-out.
C. In context of Agglomerative clustering method, ho*, to measure (06)
CO 1BL2 distance between two clusters? Discuss any two methods with
example.
Q-s. Answer the following questions:
[181
A. Differentiate between Multilayer Feed Forward Neural Network and (o6)
CO2BL5 Backpropagation Neural Network with suitable diagrams.
(oR)
A. Why K-D-Tree search method is computationally less expensive than (06)
CO2BLS LinearNN search method (Nearest Neighbor search Method) in
reference to KNN based classification? Discuss the reasons.
B. Explain the working of K-Means clustering methods using suitable (06)
CO 1BL2 diagrams and example.
C. Discuss the working of support vector Machine for the case when (06)
CO 1BL2 the data are linearly inseparable. Use a suitable example to explain.
(oRl
C. Explain the Bayes Theorem and discuss the advantages and (06)
CO 1BL2 disadvantages of naive Bayesian classification method.

Q-6. Answer the following questions:

[16]
CO3BL3 Discuss the working of Decision Tree classifier and using the data
given in the following table, find the root node of the decision tree
using information gain as a feature (attribute) selection measure.
Note: As shown in the following table, attribute "Enjoyv/arerSport,, is
the class label and sky, AirTemp, Humiditg, water are the features.
Also discuss
ss why
wny garn
gain rauo
ratio ls
is better
Detter than Inform a tion Gain
Skv AirTemp Humidity Water EniovWa
Sunny Warm Normal Warm Yes
Sunny Warm Hieh Warm Yes
Rainy Cold High Cool No
Sunny Cold Normal Cool No
Overcast Cold High Warm Yes
Rainy Warm Normal Warm Yes
Overcast Cold High Cool No
Rainy Warm Normal Warm Yes

P"Xo B ob s