0% found this document useful (0 votes)
97 views46 pagesML Unit-2
Supervised learning is a type of machine learning where models are trained using labeled data to predict outcomes based on input variables. It involves various algorithms for classification and regression tasks, allowing the model to learn from past experiences and make predictions on new data. Key algorithms include linear regression, decision trees, and the Naive Bayes classifier, each suited for different types of predictive tasks.
Uploaded by
nagasuri lavanyaCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
97 views46 pagesML Unit-2
Supervised learning is a type of machine learning where models are trained using labeled data to predict outcomes based on input variables. It involves various algorithms for classification and regression tasks, allowing the model to learn from past experiences and make predictions on new data. Key algorithms include linear regression, decision trees, and the Naive Bayes classifier, each suited for different types of predictive tasks.
Uploaded by
nagasuri lavanyaCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 46
ahs
UNI
SUPERVISED LEBRING
&. What is supenised Leavning) Explain hoo it works aves}
Involved in TE? Type Supervised Leayning? ®
t* Supenis ed Leasning is One 4 the types 4 ML in which
mochina are banned tained by ting wet lebelled -heining
dela 4 on the buis 9 tint data the machines predict the
blp-
labelled data meant one ifp data thot is hea dy
tagged with the correct ojp:
4 Ds Supenited Leawing he baring def proiided to
| athe toodel ucorks ca the Supenitor £ teaches the mode to
| Predict the ofp conectty
|g Supervised learning is the prcen 4 providing ip data ay
well ox conect ofp dala to the NL model:
ja The alm Qa Supervised eaxning algorithm i8 fe tind
\o mopping Just to map the ifp variable x with the op
| Noviable y-
| How. Supenised Leonning woovks¢ -
| habelled dota
| Prediction
\ s2o Isodax
| os o +l
| a
| A trio
| aa
Fel dole,
|
|
|
“ae]n Supenised lesring model oe -boined wing labeled
dotasety where the model leon bout each ‘pe ¢
Anta ee - |]
£ once the: Haining procen is completed «the ete) is. detted
onthe bauis qe cata + then if predict “Be ap -
gto the above. disgvm we hae detaet of. cittely pa of
chops which include: squmer, rectangles, Ariange, polygon «
& Now the chal step. is hat ye need to train he Model
with each Shape * r 1. tilde ot
> Y the Quen shape host sider -£:all’ Be: sides are equal
theo it will be labelled as GgUANE Ne noe
2 Pf he qien shope toa 5 Sule: they ik 1S @ triangle |
—> Sp the given shape ba 6 equal der “ther Hie ‘labelled
3 Hevagon ee q
|| sper Ahoining we det our model |b wing the tat, dala
| get & the model har tp identify che iil ie
| | tthe medel is hey Ayaired for all Mpa shape +
when it Find a new “shape it clayitien the shape 6n the
“} poais gre; of Sides & predict the i ee
‘lege ‘involed in Supenised Learning
Slop 1 fel detersine: the ype taping dolasel «,
- [ehep ore tailect the labeled raising dataset: 7
Jep 33 pil he! rating” dattue “int “citing ‘dedael tes t
dotael 4 validation data: 6 > i
shep Delowine the jp Fealnes of the chaining’ dalasel -
sep S*. Determine the suitable algorthon 4uy the model: Such
a1 sum, decision tea ek. Farsov Frost
sep 6: Erecule athe algerilhy ty -he haining datoret
sep 3+ Bralale the Getn¥ady a Medel “toy providing the
fest dataset, the matel pediely ‘the cored of
“aye it Mean our Mode) js ACeMate:
pe gf sepewised Leaning bon oi ply sch 4 There ave’ 1 pet q Supervised ‘Teasing Blanifiny 3
|} Regredtions- 1 Wa \
aR yesion olgoittin: ane wed Hi hae ie a sellin
the’ ip’ vaviable:-4 fp: variable»
4 9b is used ty the prediction af continu variabls such
voeather fpecauting , market -hendt eles on
some of the: lar yeqqeesion alguitim wed in
Supervised hearing axe: ' fo) od fit,
W Linear Regrestion
{OD Regrettor Tet
KW Non Wines yeqrelion
OV Bayaian fineay Regression 5 A
\) poly nomia} Reqettion- ; :
i calilctins ats lle
% Canifieation : Algorithm axe ubed iohen the of Vorioble ic
cotegorical: which meant there ave 2 clouer such a1 ye" oy
LIND, ‘male! ov ‘Females Tyuet or \Falte? ‘Human’, tr ‘Animal’,
Yeodt oy Dog’ ee
cloulficotion ‘algorithms axe wued in’ spam iberi
é mn
| Detection, aud - detection, Jinag e ‘ctouilcaton Risk ae
¥ Some af. the popular -Clawificetin , algnitboe Used in
Suparvited Leasning ave. F :
Us Rasdom exert
W Decision trea ,
vib” LogisHe: Regretion ce ov)
iw Support vector MachineCavm.)
ae , Supenvited , Learnings :
A With the, help; of: Superised: Teaming , tbe model! aan. predie
she, Ol. on the. bouis prior, experience '
In Supervised learning we can have (On exact jdeor about
the claner of bbjeck- :
Lx Supervited learning rodeli help wa +0 sohvevosigus veal tin
vould problems such as Froud _detection » spam iHtaing ete
0 Hey
’
Pisodonshge b
|e Supertited ane mpdell axe not. Suitable: for handling
athe Cornpler. tou ks: Coleen w
Re
& Supervised teaming camot
the let dato 18 differen!
Lk Training requires lot
Diffeentiate bo clay
time:
edict the conee! dato ovolp it |
tom the Araining datarel *
foie and Requecsion
I
Isne
(ossification
Regression
| classitication rmeane to quo
up che ofp info a alt ve
clowification algorithy pyedictt
descrete valve ’ ie
I, classification problem cht
Ke
cd
moore Caues
clanifieation problem wolth &
done is called Binary ehyi-
Fication, more than a clawes ig
Called rnultichuificetion prbble
90 cloasificerhon model the
onget yoriable con take a
Wiscete get of voriabla-
19k is a Supervised heowvti
Algoihm . . - : i
To oases the rnodel. Fit we
ned fo taleulale “the pacentage
of conect clossificattn:
Application + Email spom cele
inn, im09e claxifi cation 4rad
=
OV
a
af
Op
\s labelled ioto lord ov ||
Regression meant “to preclict the
olp valve tuning the “Haining
ae bet :
eon algo nedicls
| cobtnuu yalues: : P
Regression problem aeguiter the
pre ction 4 au Quantit
A Recession problern with
multisle ifp variable is called
multi variable vegretcion problen
D Regression model the target
Noviahle con take ceptinuout
‘ale i.e; 4ypically @ veal number
De is a Supervised learning
algorithm « ;
FO Aue’ the model fit we
coléutble “rost rear squared
eyror
Predicting voeathey reporks ,
Predicting Sepie in a y
detection ele-
Predicting the howe prices ete: ,
Spee tnd
31 difference blto data
‘ oeewsately 4.odfecth.
ke ee use SoMe
f)
} Eudidean Dislane veprerents the.
Common! oma 4 ML:
Euchdeah: Distance *
in disorte bosed “pelhodt in ML with erompler
stance bared methods help to determine -the similarities 4
prints, enabling algarthis to pedoim more
the Jurdomental’: distance. meouuretnen ty
chutak sboight tne a |
dione bho @ data 1
*e is-wed to ae didance bho 4 poinis in @ LD tray |
pa.
+ Euclidean distance Formula is denoled oy 1: I Cy Cyy-y
yo Cy)
dt Manhattan Distance:
te ig named otter athe’ NewYork city: vii ‘
+ Manhattan distance calculates distances plo» pont. wing
ns. *
ant tilly plait prical “‘Aordinal dota:
the Formula For Manhattan dielance if denoted %,
_ ditiy)= = big i}
4 Minkowski Distances ,
imeem Distonee once ele the Euclidean + Manhatian
Distances
gak is scoltulated. on -the bouis 4 pete porometer’ p
{ ithe formule. denoted * Minkowski a
} y * Ca hen I"
P=
( $-consine ei ori
| csine cosine, sol
meoavrel the: ion ‘bh Jou bay
her le ee Je: SIMil uy lw 2 veelors based
‘ahis porticulatly wed in, teas Tike: text Sinibnly
Lfouhe’ se sat ty
Conmsine Similar & rectus ig fol
iG the congle “yd chem: ; sued | ua “eco Sim (418) = (os(o) = AB
fon BI
y azo
A
K
1S: jameniog: Dati Distance:
4 Hamming! Distance it’ wed with ‘Binary date: ditfen
4% cokutlater the distanee ty counting soy af ha ers
“| bl two strings: «
¥ This plays an inprtont dle in dota tronsmision A snot
conection coding
A [ap fluke EN
'
8 De Ts jarelie]
Hansing distance 6/B)= = _
| Whol is, ENN algorithm? explain with an example” (1)
What ig R? tearest le neighbour aloof -
AD is boved om Supervised Leow ie
HRN is tued: for regrewion..4.404 a to chusttion, poten:
Bub sit mostly ES owitid ton: problens
£¥-NN al gorithm ‘Stores: all the ayollable. dato. 4 cluifies new
dato. pinks ae on -the similositics
at Means’ that» whenever a nevd dala pi arial it can be
clewitied the Suitable , corlegeny by win K-NAL algcibmry
* rk IN “ig a non-Perametric, aban that medny’ rEdoan't make
Tony © oxvmptiont: on’ ‘underlying. data,
leat i ernie ou hax We ‘Nloorithny: because it doesn’
Jeary fiom. the doin data set immedialely-
Ht -NIAI algonitin in tats lage it. st stored the dataset £ when
Ho. point aniver “ther it cloitien thet data, - & Cat
ae : civclef 4 ty 9
Exe Let ud tonsider, we have imac
‘ptor the! identification #3 imager ube Kean we KUN pit.
> cadegory - A-clrcle
tel : inages ima BD
= aa -Oo
taleory- R- triangle:
training imag es: AA
A
| R&R
| A
BThuc Judo datouet iinager are tained. with ‘kon algoritn
[40 new dataptinl . astivel U-rewdata point then kK- NAD
tthe new coda points) alarm checks or the similaistia 4
| identi thot this | rab point Fou more Ginilnibes
a A-cireley 4. this dato. pont | ig -caleguired 4o Ae
Grol ical Representations eae
|
|
| AD, a
Yap oi
pro
On. pe 4
| 0 ory -
| Le gee
we need RENN algenitl al
we ‘Bae ond Fey quien yA + eqOYY~
when we have o SE point “the onl dditn
“puahe
i Find which’ category the ol belongs is Ken valgibo
| Hoo does, KAN. alos =
ashep It Select te dhe numb ‘K of “the mrp
-/glep tecakulate the, Euclidean distance - eh numberof hen
ep 3: Toke the KP nearest nel Ibour a8 per Euclideady won
Step ys Arnona the zt neighbour! count the ep: Lo ets) uh
skep 5 * i the the ae dato. point to that one tor
“| wh) he toa ne eighbourt 1 Moximun .
dep 6! Dutta re ApheB Hg nish
on Qomndete
Ayn Pow
Dyed
Vey cs 8.008 cate
090 a i
Before KNN x # Attey K-NN the new dota point is categorized 19, Coleg in
: MYA
| Euclidean dislonte: 4. fea (ya) 1
SEN RTC sis
: » Alu)
| Kooy) oo!
| ow do itd ate, kia in kcal got:
there is no porticulay ody or; omula, 40 determine the bert
value ‘ ,
need Jo, try, all, the valuet -lp find ~the bert “out y
SE value must be odd numbers,
Sg ig even thoe 6 a chante a equal feoduves
Kean the
Hy 2 new data elemen! to O
ly MY tbe value 4
in cssication - So. that vise cast, low
particular category: er ae
el, RN pent i ‘
be is! single to. implemen” Be bilo) a He rag
ROL ie solu 40 be coy tig de sila lenges
lx can be more effective iP the “boining. fa. ig ong:
ieodvaritaget ap kennt: o wngt alr anu
Disocion) 1 determine the, value 4 K &\ somebimes
Heth is ves det
erik is! y Gpormplen Slog at yer ibd ott ate at
Te we het to F ieule the distance ble oll the datapains inthe
Haciring dato. whieh is very ceypensie fine comply eoeaie
"I 4 eh -(gla } . ie
S
"shape - § erica!
| h
rye
ji
|
|
WOR gg Le Le
hat ie Nive Bayes algorline in MLf Explain woth ao Cheng
Naive Baye! algorithm is Cloujfication Hechnique algority,
Nove Boyes iso Supervised learning ak ithe ich is
hated on theorem 4 wed for adbing ‘cloutitication problen
# MN is mainly wed in text clouidieation & includes high
dimensional training datos 5 Olositice,
# we We Ata B algoittm whith is a move elie, ;
alguithm which helps in designing fort HL model that make quick
[Pa Seep clayification al
| Pediction are made on the probobi ity a eo
jk Somne polos examples where Naive Payet algorithm is Wed
| ae spar titer ert clouif ication, sentimental analyis “
vBiby eas Celled a Mave Regtt
i; Naiivet ot +s colled Naive hecuue it assumes the ocourences
of a. certain Feature ‘is independerit of octuventer | another «
| persent ¢q
| Feoctuye- oh egal pecan
| ext ba fruit ts identified on -the. basic ¢ colvsshapeptaute.
| Zeplor-red
| Joute- Bppldtavte '
igcthen if ts rete Apple é
[A Here each deadufe individual contbuty 10 entity 4
“ik is an npple without depending on each other ‘t
| 2 DE is Called Bayer Become hit algocithm is bated on
| prinniple of Baye theorem” - ‘
| Bayes “heaven
Baye “Theoran nalio Known 01 Bowyer theorel aule_or
law: which it tued’ 40 deterrtire the pokobilty
pothesis (Selecting More impatont Feotwa, tor por
pair knowledge: QW dependt. on the Condition) proto
The: formula. Fov__Boye “Theorem ie given OF
p(B) =P p/n) PO) :
PB) i:
ba bility: probability 4
whee Cole) — Posterior
it Bon Bethe obterved event B
p(elh)~ rdetitnod probability probability of the
exidente Given that the probability QF rir ES
k hypothais is: Gi hue:
Plr)- Prior proba’ ity ¢ probabib t other's ptebo
observing the evidence id ie : ad
PCR) Ma inal prolmbiltysprobobilty 4 evidence
\wlorking gure aye clanifiey- b ba
Problem Assumptions: i
t-we have a darasel of ewaather conditions on corrdiponding
torget variable “plow: :
Ry ul this aseunpion woe need tp detide thot tocarther we
should play or not ty 'o porticutar aut acceding to the
tet pollen wwe need to fofow “he belauo sleps «
dep recorvert the given doe set into frquey table
dep a8 Generate iletbood table by sinding sprobabilifia af
pen Fecrtuver : ; i
dlp 3: Use. Baye thee +p calle postaior probably.
Dato. set: Piel pea
[swo outlook [ph
| \ Rainy | aa
ye), gun Joye
2 overcast | Ye.
g, ., |-overout yer
Ss Sunny No} */,
6 Yoiny ya
+ SuPny ya shes ae
8 overt Ma. ,
1 gy aa te:
Jo pa No’
{1} | Sunn ye
Ue, Jp vaing 7, Mo.
13 | overeyt ya ||
Ny. | overeout | Yet Julab te a 4st!
Frequency table tor the weather conditions) any Te
overt 9) & | ‘ rfl
Rainy L :
Sunny | 3,
Piofapptats if tO |e Ukelitood table for x wpearther conditions ¢~
"povatier (No [Yay
parat 0. (5 | bhy=o3e
Rainy |e jr | ylty=o-29
thy Raye “theoem’
“plyes [sunny = p(Suory yes pla)
p{sunny ha) = Sfip =0:33:
plsumy) -p3e
P(Ya) = oat
PLYe [sunny ) = (0:33) oti)
Dar
= 0-664,
| PC Ne] Sunny) = p(SAMYTA POSE) peg) 2
| P (sunny) ~ plied =009
| > Roxs)* (028) Pleumy) = D3
| O3e
| P(noleumy) - oul
| Hence ik ic taltod pl yerlsunny) > CNo [ sunny)
|
Ix ence on a. Surny doy player can, play the game
| Ad vantages * : yet
| gnlaiver Bayer 18 one of tie fost 4 cai ML algorithms to
| Predict clow of dotasels ‘ .
1 DE con be wed for bay os well cy rutfi-clu clayifi-
| cations: . f
qe rite’ lov choice For text claniFieab’on problem -
1 DL perorms well if mutli clout clasiHication .problenu
| compar +0 other akqovith
| Digadvantogeé :
| & Nave Baye ouume thot all Fealores ave indeperctent of
| Agclnted sf HL contd kam ‘the. ‘relationthip: bleo -eaturer
nor | 2
Applications $ Native Bayer classifia
+ YE is wed’ for cre "cot
4 DL is ted in medical dath clauiFication
it GL is wed in Realtime predictions :
DE is Lted in text glouHication, Spam FiHering, Sentimental
Analysis \
Lsthere are 3 dypet 4 Naive Bayer Modell
\ Gaussian Model
& Muttinomiat mode} i
3: Bernoulli Model * >
Define Deckion -hee- aks a STEER PPE
buling 0. Deckn “cee ul troining & visualizing and
Define the Follaoing teams: ti) Entroy
UW) Snforrration Gain
Ui) Gini Ratio
) Entroptiy:
A Dn Mt Entropy mecuiner the level of “impusity w disorder or
uncertainity in a given data sets
Ie DL is af meric “hot quantities(estimate) the amount
infomation ina dataret & it ig commonly cued to evabierte the
quality qo model £ it's abllily fo make accurote predictions:
4 Ai ay sole indicoter "oe ures] low eniiopy
ue indi Igh acenvacy in a ML moe’
ee ‘1 thigh eo Low entrpy
&Mathematic formula for tobopy is 9
ets)=— Fey! Roy Fest es
Tar @ yeiCave) outcomes
otter splitting 5 Yel and
a No
Ex Lel us consideY 0 Feature “tho
ond & NoG-ve) euttomes : dnitialy
3 no & splitting 4 CMe outeonel) 1 3 Yel 4 feaduved i '
te -(& Jloq(£)- - (4) leg Gi hy
i £) AES) abe Ose -O-88
= 11584
= pag
eet CD ble) -Ceald)
MBI cola)
iat Ti
log
: ceuntauy 7 EL es
=. OWS + 06206 ;
=0as43
| e(eaines = -@)lq(%)- (%) log Cz )
te es('88]
(oq)
- 06 (0 ae (ou) ran)
= pure — +018
E ro 2 09t
ci) formation Gain (i 1G) +
sp Information Gain is a, measure how much information is
eins by splitting a. get of dato ono. particular Featue:
4 QL is caleulated boy compa ning’ the tnbiopy {te original Qn
doje. tothe Enbsopy: eae: two child, sel
A Sie information gain value indeatel hot she Feodene
ie more eFicient & Heelve adler ¢ pit
“ete formula for Tnformatin Golo is qren by
| rq (se) = “ )- (go
| wee TALS) ya) 1s the informtion gain tox the data set
{ «+ the Nariable “a! Tor". random vor able Us) Is the Enbiop PY or +he dat ty
wcle) i “he cn ery the dataset gireniche
voriatle °o
Enbopy -@p e(riae ) Gate ara (Z)#0m)
ganar oryoe! OF 40-38
Iq: el) eC PA) ‘ons
2 lo9r-0-98]
=0'0d
e(ti)= — Co)log Ob Cb.) (24) fog (5) Ly
ae » [S 5 Cf]
= 0538 (0:I068) - weit taon)
ae 7 + 05I-
eth) pets” (blag (%3)
Ee tgs) (is)
cs [S81 @[ al.
. 1 E Poms} - (0 040)
: i ie x08)
et) - U8) Jig) {ug 8)
CD SI |" a)
eps!) -(os) &1)
|
Wow time
ECE = B)voges (Bat
= OU ROI + Orb
rorrttrore aoe orb
aObe he
14 =| e(P-e( PID]
= lo-4q-o-4
208
Design Te’
KA decision tee is defined a the Supewised hearing
algoitthm ured for claificetion 94 oxll a yeq ress problemy -
Soul tication problems:
$ However, Ht Is mostly wed for
SFL shuclure similar to a tree ushere Infernal Node
Yepreents the fecduret of -the dataset, branches 4, ibe ree
yepraent the decicion dytile, 4 leaf node seprete pe airkcomel
# Detition ‘reer are wed +9 predict an outcome bared on
historical data: :
& The Dxcigon tree eoork of) the ‘saquente &. iF-then-ele’
gkatement £0 wot udbich is on initial problem +o ole
looks lite +
a The Decision tree
"ase hale
\eof rode:
| leat node is defined a1 the ofp of the decision node
| they do not contain any branche: Thod meant the tiee capt be splrrtecr Furcther!y 7
Root rode’
Poot node it the osigin print of any Decision tree 4 H is
Ola eet hich ee ie
4 I contai ive dota se i vided i
fas gee divid Dh Or ove
47S rode inchactes rvaktple ‘branches 4 tt ig wed to make any
decision .in classifiecction problems /
split : t
* Tilo process that divides ‘the toot node toto mutliple sub
nodes bated on Some defined conditions -
Branches:
# Branches are formed by splitting the rot node or decision node
Pawuning + ts
x pruning is defined a1 the proce Yesnoving unwanted
| branches rom the tree:
child_node* : .
HK Esgept the root node all othe nodet, are called child nodes
‘the deci tvee-
Using Tds_Algovbm * ;
bed aoe ain vx, build a. detision dyee ty a datavl
£103 algorithen meant “Dlerative Dicholomicer 3° | stnvol outlook Tempecotine [humidity | tind | Reitan oe
i [wot | high = fuak |
a hot high | hong | to
1B frovercart | bot high weok fa
| rodny mild high uoeak ye.
S forcing | cosh... fpgtpal | WK. | Ye
| 6 roiny Go norral Sto No
+ | overeat — | Coo normal clon ye.
| S| sey mild | high wok | No
4s} sunny Cool ora) = *] weak” | Yes
\ov ‘cing mild normal | ora] Yer no
1} gonty | nig Normal | Strong | yet
Wee | pyercaat | | mild high |.strong |yet,
| tar] overeat | hot normal — | uoeak™ -/Y¥es,
Ly: | rol ‘mild high tong | Noo.
| pnsidey the Firs the isk feolae il if
ya" | =e |
pe 7 Ss
e q 0 g
Rainy: y bos
iid Sviboy le
entopy atk
Eloutteol)= - Ry Paha ~ % log Pee
= (Ga) Malde)-C) 000)
a) ST
=~ lobe) (-0-65) - (0:35) (4.48)
20:43 + 0S!
\ = 04l
“Temperotre ya Py Toto! yh
hot te 4g
mild 4y r L
cool el in
4 5. Te etsunny)=Le) loal &) -2) logl 2
=-AeSeg] -c2) [92]
A ogy 09
= fou) (132) - (0.6X-0-43)
= 0:58 0 UY?
= oUt
Elovercart)= — (log (2) log (2)
“a
z io ()-o
ECRaing) ~~ (3sloq(3) -Ueegl)
--@)fig@]-C]
2-(09 (9-436) lows)
= OUR xo St
2th
ECPHD = (E) aoat) x0 4 [He] 1076
(039 20:44 4 02KO x(D:39A0°°6
epatyer vx 034
204
Fafornation Gain:
Ge ECP) - EH
2 [oa-0-64)
dep ts roa sl os shy Neale: Temperine
| Tempevalure yer] ae | “rola
hot a Lv ; 4
Q fu bo
a A) 4
af el we | Eniampy “Jemperodue !
E Crepe) =F) 4) eat
Gall) }- Cg]
sate tedon
Elbol)= Lily AB 2%) logl 2)
~ te) [eae] (2) flog ted
# ton Let" |
= pay 3
elnild) = epi) ee
e et (: a so)
2 pe
eferol) = ) Teg cat)
2B) Jeg] - - A) [2 |
2 -(b49)Cour) - (& ey ;
=O dt OS
= OF 80 ;
ECPHE Lata (x0 ¢ (4 20¢0
= (orig) + (o-w)0'I0.4, > +1) X080
= pig 40-348 tou
88
Groton Epi0 *
&=| El?) - ete)
{o-al- ol
O03
veiw Sep 3% Let Us consider’ the feature: ealarty
ony ea otto!
High J a fy =
Domal | ¢
S
+
Enhopy ‘Humidity: wd]
€ (Humidity) = (4) 9 (82 )- 0% a
= fo pig]
= ~(o-bu)(-0°63) - oe
204)
eGHgh) = -(Hleq@)- “CG
<2 -(3 "Ge!
= (owl 122) -(o- sao go)
= es a bush
€ (normal) ce A nt me 'oq (3)
= “te ad 4 Jpeg
py. cet
= -logs)(ron) - : Wt @)
= D184 FO39L
= Ost
c (PIE) = Giq)cob aE) x08F ~
= (os) 20% + (0:9) 1OCF
_ 084018
=ot6
Sntermation Gain t
1G=|etP) - -e (|
2 forat- 0-6]
e ols Dafertrotion Goin t
a "esl >e(e)-E(Pl)
= OU-OFF
Enbo
+ (plr)-
end“) i) eal 5)
Sep a Let ws og dhe Seature+ lind
Wind |
Weak
i |
—j
Winds
~la)| re). ae fog (4) bald)
= -(o mate 63) ~(o° i es
= 041
e(Weak) = &) log) -() "9 (% 3)
= -[eyfr iP
a) DI f)- Ca rls
= (or te) bo 1) -to28)CD)
= OBIS
e[sbrong) 23, bsg (2) 2-Io9(2y
2 ® s
= &) | J-@[! ep fa))
= (0 en. (os)
Seatos
a ae eida B40)
= (ost )lpgi+ours
2 OUbITOULA
en)
=O00L Sep awe reostame -the datoret. ox shown belt:
loutlook_| Temperature | Humidity | - coind | yet/alo
Sunny Ree J high Ly weak t n0
Sunny hot high sheep m0
Sunny mild high weak no |
san Coo) Norma} weak yer
Sunny _|_mild cnewal | shoo | |
outlook “Tenparatiare [Humidity | wind [pals ;
overcoat | hot high [strong | "yes
jovercout | coo) novo)" | hy
excaut | mild high | weal fet
vercout hot nomol weak. A
outlook “Femperoture Humidity wind yerxlo 4
Rainy mild igh TRAX
Reiny coo} normal, | * week :
Roiny Coo} nomal | Shog 0
Rainy mit ian : rises ¢
iny n i
aaa hos “hgha Jatorrration Gein C1G) when compare d
fo other g0 i 18 selected as parent Node For the ee
Ly he diognam hooks” ke -
[outfook |
Sunny weret —-yoiny oF
)
) !
Entropy omy
ElSunng) =P logheyy Res Fey
eehg eZ boge
2O4 poe ob get .
20-5240 &Y
e (Lemperoture) =O 4+
E (temp=hol)=0 Nd). Llpg ¢: .
e Ciemp- mild). ~eolge Slog CE) he
| es e ee
vi | edpetpeatitd i
ECTemp= cool) = ou Jog)
ee ac
1) eLparent fates prove pater
q =6.
Tqe tenn)
Tobe e
| oe ony
a by= oF 3 log 3=0
eae =O |
E (porent /Feature)= = fxorpoo
T= eee etry
Pow 0
“Elusind) = oy + OAH. « ue
Eloathe “Ay de lty C4)
= Co- aw) f Py (oo ate
ine ee o ay!
e (sh) = = bg @)- hath)
\
elperentffeotwe)- E(P)-€¢P/F) |
elpirj= enm szK0
Iq= a * etoe)
i
{|
hi
i
tid
Aa
bai hight Ierfomadion me 4 ibis get) 01
And no look (usthook 0
Sumy. , overcost iy
idly YO,
| he nb : Enbopy (Roiry)=- 2109 (3)~¢ hog’)
2oqt 4
Elfemp)= Sey
E(rild)= 9% log,
\ i om ‘
Eltoof) = 1
14: €(p)-e(PlP)
E(PH)= 2x04 % Yen)
2d: ae 4
Tqsoo] «=; EC windy)> 2044 Hd
-ethumidity)-09F elweat)=> vod
ethighs blog} - +a, else)
lv = Oo syn
Elneumal). “Bly ete ESC at-D pt
he
even 20-90 sls i
TG=0:0) a
ihe Final decision -lree cing 0 algithn jutlook’.
{cuitlodk |
tice
ee Pe.
“ig roime) sey Ste 4
pio ea, oth BLM | 4
Expain cart algpithm or building secon shee to) “this ang
¢-danification
a- on ;
i Panne ;
epee alqpuithen decision awles oe "ind tl fea dey
K cart algorithm is kno os
“ios al ine) Humidity | wind | Phy,
[snio|__outlok sotuve midity win
FY] sunng ae Tigh weak yaloo
© | sunny = hgh | 3h © |
3: | evacoit hot: high wR , a |
H]Q% | Roiy mild nq wet ja |
Hage | Rosny Coo} non ook |) 4d |
6: | Rainy enol Novin eon TA |
Lys |Owvereat | ceo! novel“ f Serorg | PW
S| Surry mild high weak ny
9: | sunny coo! vernal et 4d
‘Hor [Reiny mid bow uoeak ya
[ft [Sosy tril rma | shang al
las lovewdt |. tnild > |, nigh ste da
13+ [overcoat “dot normal ya
Hy: Rainy mils) high cleong nO
| xn 9 Tosler. vale
ig
redotl Yrder. is a mebie br clauification tasks in cart abpoitha
ADE gore gush "4 squared piokobilities each chy -
$ The Fomvla for! GIN wiht 16 gir
p Jor j=) -fo oe cong
Sep saint 9 hdew value for outhok. * .
outlook ye No ihe
Su + 38
pvertas y o
Reiry ‘ L 4
Sy
Gilat) + fy wes)
8
Gin’ (owtTook = eee Less Coley | 4'n\Coutfook-Rainy= 1-252)
=0:G8*
oe will calculate the cagliea sum of Qin inder-br cutlook
feoture:G (outlook) = XW G Hore 1G = 0-34
Temperature yer No Nov Moices
fot LS
mild “y L : I
cool, gy 1 g
Gini Chernp= hot). 1 [CaaeSleos
ain Coane fap Ce ore
Gini (Lermp =coal) = 1~ est
|oe will caleulele Pee ea Gj index for temperature -eatore
| GiLeeroperatve)= fo te §vorgury x037 SOMO HAO-1O> O19
| Hurridity yer No Nog Stic
| High Saniee 2
Norn}, c ay a
| Gris atomidtysigh}= aye ey-ou? \
in COuovicly = shaeng) = | Wyo ty
woe colll coltalate he: weighted dum of, 4A Inset ur dy»
| Qthunicil) = £ so yarmly -036
| Wind ya vto og) instance
is mek 6 OB 8g
gig 88 a
Gini woied weak) 1. [egy CO Heosie
Gini (sind = shron) = 1 [@dCeyf or
Gsind) = fg x0-34 hi res Hod DL OGL
Bere Qlouttook}=03
| Glenpectune)-O°8
obit); os
‘ eb ‘
the wee for ~he Decision “tee is Sekcled on the bas
|
|
|
ai ue: value Alors parent mde mut have [ovo Gini inder valie.
0 D 1 Gini of terperature-br sunny ottHlook So
“Temperature yer No Neg instances
tad toHet shal ib. in L
coo] \ o ee '
mild \ t L
Gin (ourtook =sunry -lenrp= Hot} 1- LES (JJ=0
Gis (outlook » Suny 4 temp =ceal)= 1 C4)'s (9)] <0
Gin (ovtfook = SUMY4 “emp =mild)= |- [esc
20
we will caltulate the weighted sus 4 Gini inder’ of
Lempevadine Foy Sunny * : f
Gini outhok =sunnyd temp) =PEJor(L)p epee
Soy
Gioi__ Humidity of tu outlook +
Numidiky Yer va ol ln a
Nh 6 8 i dig
J Nima pe OS Tera
| Gi eutick = Sor + umily -bab-1- [eefs(4y'T =o
| Gin Toustnek = Sumy 4 Humidity = Nomal] - -[)"- (2)"Jen
woe will caleutate the welghled Sun of Giri inde of Sumy
ox Sumy 4
GiniCouthok= sunny 4Aluridily)= (2) x04 Ceo
+0
ini of wind of Sunny outlooks
wind yea wt nog Snttances
ork yb 3
simg \ 968 < Cytibinedt)y
Gjiootak = sonny windewead)=1- [C89 05h]
Gini Coutlook = Suny 4 wind = Stverq) zie fay. ays
a
we vill qaltnlate the uxighled Sum of Gi; Inde
wi nd) for omy - : a ‘“ t ‘¢
“| Gini outook = surg usin) = Bo(eer
© OFA4DG
= V4 | Humic
te
“UL eet
ain
No
No |
No|
reetukleees |
aa. iy
Tttumidi
a, . Pog PM PeLEem fra gona
wf Ww | Mes
Ny fi : ' is ‘ w He |
Nv
I ¢ v £ [ho |
| to we fw {Yee
4
re PP
zee
ey
| x {. 4 g | Ny
Gjri_oF Jovpewtae eo youth
3 Temperate No No- g instances -
uid 4 \ 3%
cool \ \ 24
ipttok = Raa ompenodel= | [ee (APJ 0-44
ear Fa kg eob-1- [ep iC -os
Giotto Soa wwe
oe