0% found this document useful (0 votes)
95 views25 pages

ML Part - 2

The document discusses various machine learning algorithms and techniques. It describes linear regression, logistic regression, and decision trees. It also covers metrics like mean squared error and mean absolute deviation that are used to evaluate machine learning models.

Uploaded by

Ashu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
95 views25 pages

ML Part - 2

The document discusses various machine learning algorithms and techniques. It describes linear regression, logistic regression, and decision trees. It also covers metrics like mean squared error and mean absolute deviation that are used to evaluate machine learning models.

Uploaded by

Ashu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Date Page no: 99

4Cemisbi
R can he sedkx ophnizin h dsamicalzeachin

5uusineSS

R i s no ultod k_business shalkryy-pluminy


6. anu fuctalmg_

. n vanions aLutamobie manullahinng Cempanies,h


yeha se dsep heinlecemet leanning pkk gsod
aue put em n Jem_CMtunrs.

inante ectoy a
a R L is uRLenty tused in hinante Sector.
Pr eualaahing kadik shahsies
SC

tniaton oKegexion 2
R oxinveaigaking
ha lahtonskip Gesk&nndegen dent
**
OYeawus n d a dependlentUaiasle
awteme 's wd asamehad era
b

madelingin machivleanin nhich


oOhm ed reolith ek&ntinao
autuo nmes.

h o sJachinn leaxning ketueAATOM

Kegveiorn S [Link] shtiis


which oms a k Pa forecosEModek
madhine Jeanin
apeoack Prole sondinuau etcomesn
editktve madeNig Mackinedeanin Ra
2xall nvoluesplattinj.a l i

tan e predicedonce zeladlonslip leke


indepemoet aa clapenolet vanuiasles ilss bes

uvaceggnss namn
eqre ssfon Mokelg e

ackine Anna- Aeasion madels


Mae
Manl de redic Alecmal
w uemds Ad Ytelit _ottomel Hesred
modelsill e ained tunder stand
oclodton ship Geee sikleret iadepmdeut
ate Page no
*

Vartables nelA an
cukemes: S s CudGe
prealiet mssing isie data sa
n a 3alea enui-
-uanment a
eclin lusaaing rsdie Paltet manths

Sales ro .bey y_ o achus maee


Made

be e .ainee n laGelled pk uput


ele unalexstanal speile eaturel hes
classi akelhas jeck
proGle Jhn wsclel is
Prealtet ContinuG outome Gr vaws.
S L s e d be Mede H orecashsala -
olanaes e pice a Lehil3slea. Cammtk
FoY machinw eaing e M l o n M e l e l s
inslusdes aieHa ucLe uha
e i salea ax markehlna anapaign talysi
douses aaGlish xelakionsbipá Gekee a

eetthn inlk seat aea shckpices Ban


aVode Pachurs eaxdtuJ [Link] a~uali-
SLdin

kezneim.
e e C Adne ol edapproacku
h tereutekaiqus
ae Ierestnumbexsol independeat
Date Pageno23

**** l x e tapes Mackire lesing4tes


edek
maclelsmay
ma alo agume diEtet
datase Gelhseem inelependent
eepemdetNaiable
ome o
Some d mo comman *vAsi
doniqna t e gaoapa HHlu

LineaY [Link].
MJieeA yesmeata

lesis
simple nearKeaeaim

Nimple lincarheyreim isan a


giple
Dplati skaight
ch Pal n imise
eOY Geksee
CemnMen Simple e
eca af tagltli

iple linea rsoaan


AWkple l a r veeasim e cdhacaa
e ukam
Date:L Page no: SD

vauiad le is e namia eexiois a ye


eleliaeas egpeitne sed
C ialeedeat_Naiadle

simpla
plethe

Logis reesion

e pra kaoAi
ee sismod UUVJe

h eladisaSi eAesdens
Auelepeadet vara6 lex henallaH
Ltput Naluu
Phgrnoo G8

J
nomia [Link]

nomia Hete ion. A9- a lorme LneaLinoa


Aegexion Knson a specialCak al utol
o
iiniar
nar exton which estimaks elaons
Aanndh [Link] 1olanomia Reste.
Polnomlal._

SSfon is semsifvee oiexS S he preaense


a e n q r waoud lies Can aka oadly aP
pexFormance

[Link] Meics

Kezaenfon edictive Madlelltna e hose prl


lemg a invalue pxealicAinga numeric valu
Hebicss xeeaioninsabe_calealatins
n erOY Store_to Summaize H prealehe¥e
skll A male

ean Sauaved ror


Mean uaxeal xro JASE ghol, is ca
Paalp
aarr e o rmetictor uekenimpcklem

4salke an impaxdund less. anckon ar alp


thm.e oyapHmized uing eAteat squar
hammina g ezhedionProSlem. Hereleast
8quares efes do minimiziny He mean quax SquareL

eor sthueen reolictioma_and expecko uau

M S C s caladaded a mQan OraveyA


Datc: Page no: 6

3quareddikerendtencea euween preahche ane


exppcddugi Valu &n a olalasek._
MSE4/N* Sum for i NN -i-hal 32
lhere i hexpeckd valu in thu dadasel
-eRenee hueenHie buo vaksss is Squaresl, lklhich
Aal ald_lhemovuing Aigh, [Link]
a paMt kwrvalue

Squaing a l ad d_lked el n tin


nnagniaiayisge exzns Kal ishe langeh d
dPlerence betwees Ha Prcditled Clnd expetkd values
M larges sgAasacdmkiHiveenoy
is had t ellecd al punishing_mlels aas
kar ahgeh Lehers ulun Asg uted au_a leys
iga n Elatrg s average extox gtare wheu uiday aa
melsic
We co C caukea_plot e uim_to houo
Lclan
dLa_clang npredhtdion hhoy impah H Squaveo
Date: /Page no:20

lean %sotute 1lefadfon (AAN)


mean_absoluhe_olauiahon AD
da-sef g he doage alisance bekiweeu lao
da ant_alhu data-set and HA mean
o adaka iie_iepelenis HheaMouytO
vaniaton ed_ccws lo mean, vaho _in
_clateclakset. Tialso Meabuzg e_variafim.I
is_caleuubed ase aueag0 al um o
dh absa feue diLeaehre Getuen Ch uulus
O/Medaka-Set_auel_he MeM.

AAD (2-mean)n
there 4isn _and n_is number o_lalka-Pone
n _deue- Sek

Ez:h_dala-set is 11,16,1is,13,12,37. laleasa_


Hs mean_a6solube deviakm ol tus given_okta-1of7

olutdm

p Caleusasng
Date:// Page no:

m ean alAgiveM datu s -5


e2:Calenlatng da ahsolube_dilo2Lemte Leheleeu
eah olata -p6ÍKP Od_pmaaLn

Data-oind solute ALaenee rmmean.

11

42 12-15 3

1S 1S-15/-d

4-36 2.
9-4s 2.
18 8-4s=3.

Slkp: Aading- He a absalute 12:eremte Les


-mean)= 41.34 0 +2+ 2+b
-nmean

ep H Lidtng ts Aum_ol abselube eilleence


and Hhe numleu oP deuta-painfs.
AAD= ( x-meanl)n
uyo-yp
Pouo

7HT

Z ou 6ed
Deciston lee and Kavnclon_lorchf Classilee

[Link] (Llang_Nins afal.,200


Cheaks t clasikhtn madel h [Link]
ee aoh eede in t tec specile a kt_an-
en_aihk ach_branch dlescenna kom d t
nade coratanekt
t _ e _A sleuahes
. utibuh

eson fees SaJugen-tseoMackie Lesany


Plyitkn at wMA a set a ulea o make
cle isiond3imilax o ks humanu make clecihid

Une0 4 iak ata a i Leanin


abikaheaalgeyhain at u
make cdeusmd.

anclom orestclan i/ters

Auandan keyeA 4mek lmaor ut 4


ALmbena_desib he cliKee on Vet
Sub-Sauples a u dataset anelus avea
epve to rclehtec dcuay anal enbol
Date. _ Pageno 2

Oven i .
Oven Sub- Sumpk sihe tolle.!
wilh du
Maz sa12ple a2aeki6aakhup-1
bodstap-+
whule oleult *ukd
Laelauudt)dawik
GuldQaehteee
Date: L ageno:_89_

nkoahuck usinae
k ion ee
Decisiom kee s mo)? sed uded alsud
wln coats Supeiiul deerning khuale
T pmah Mahlyattibuks tatindiele nkmal
nedes brandes Qhel a kmind hode Ts
algaitns are2rellkard dyu h shbl nd
seliabl
algev hr predickiue analyuls.
Ho es erisicn ee eoth or

om & tu p e o p 2pu
Ats cakkaa cea Comtinosla
clamilitht ehlemm desia see
ks qultatms Cael baed

ub brunbhe. H cum ebo le ueal


elt weie s_a k Castunmn KUV_hus
inalvielaa! wa
im h_anK
ha eel - Lotu
_deLeat 0Y note
eeremetriein,
.F
L
Y tomama .AeasnreeJebtca
oClasscotcm ceua
Tis mast ComM o- 2Rrtor mane metrt
ckassticadicn dlgovitams. A may Ge dekdnad
AuMben. a conrect prRdichluna made
ol pxedirtions made e can
eslycalalatr î 6qonkwftn Madh t
6muWa-
HeeurutA- TP 41N
TPFP+ N+TN
atc Page no 68

**"***|
uncifon
Anle CUM we aICuYcACy [Link].
Skleuma metytco to Compute aCUYA auAr
elassiLcatbn meclel
***"

**
ontusicn Madix x

is s H easist o measue thpex


tamance a aclassiltcahion lem hare 4
uhare
Cx moxe ape a l clases.
conkalon matix fs notingGud a le
wth tuoo dimens fons uz Achual na
r e a i c " alno lnHay mere Gath H alimena
u 6stves IP)" u Negativelw
False PosidivesEP)'False Negative (FÑ) ", _

dual
T iesdive False asihe
2eoicdd CTP)
NeaHve
(TN)

S Yosvea (TP) -
I s t caeuken Goh atue elass 2_pxakkl.
elass olata aint 4

Negahve (TN) -
J ts du Case w n bath octual clos e Prealickd
Date
LPage no:
9

o_eska ám is O.
class_

oalse Yas?kve (EP)


T is COASe when octual claus ol
Qoint ls O pdichel clasu ot dah daa-
paint is 1
D -alse Nesalvea (EN)-
s H
Case wn_ actual olass allaa
pontis1
12presdtches slass aola s pointis O.
leCan we cankasfonMaiix uncdion_o
SKlea»n. mais sompale tnuatem Hashx
ol ourclassiktceadicn modes.
eciSTon

yeistnused adoccumen uehriuals moy


Ge_leline ou [Link]
tusned o mcdel. lale con aail
calae
calelake Cankaian mahix ithH klp
a mula
TP
eciston
TP FP

Kecall OY Sens i
Kecall maGeolekeneo a t numbe o
Date

A posiHvesveunedG ouxAmeel
le can calgulale 4 contudion_md
ormala

deral= TY
TP FN

Precsion RecallTade-ol
mode ih delesh iu shap Jh_in
mol,again aim oAo Mocles fo cla
Cu tmeY a shapliPhy_wn achuala
Shup dil means kigh pxeusion 12]GEp
And talse _osihve is luo -

mgdelhich dehch a Jpan App hcand noi


dealhr aalnaim ot mode is A
ulaa LTP/LTPA FA} T mosl delech t t
oLPplitant is net adekeltr so applicant mut
Ge oletauhr [Link] sheulel J0oluce a
uhtch ncrease 0cal.

Ontox una ctun haye Got_prersion o


Joehl Wh Heu increase recsion l
Joeclure ecalanc ufLe VeRLSa TLis s colled
H xeaision ecal aale o
E Score
,e

t amonic meun
S Scoee
precisitnand eca Ata malcsJy4
Ssoxeis weishHsd avexaa h veukio
cnel [Link]. TheGeal uclwo d e
Ge O.\ule ctu lodste-
F Stoxe lp of
L2*preision t D LCpretisicon e c l .
Sore haing 2qa 2elakve cn n
e asicn

AUC Aea nlez Rac wmue:


UC Avea Under (uzueROCCeeivcedta
Chaxakishie) is a etermanteMetie s
60 Vaaing hrehoblyalus r clasaiktcaon
prolems. name suggeah KOL k_a rcsekali
cwveand AuC meaune Aa 3epaaltl2Ta.
Simple ds Auc-R
w ad Cpaapail o mode in diHn
ceaes Htg AUC beltea msele.

Matmakcadly T Can e Ced aelaHfna


TPR T losAive kale) i.e Senstiui
sece s FAR(Eslse osHie ade) ie. spet:lt
aiou hveahsl vus STLa uph lo
RO UC aina TPR o c i a s a FPR.
cut- a t s .
TPR ROC

AQc
Indoduchi'n ttas der NeusANa
wo KKs

Neuua Nehoks eaimm coL


C G a n HunchtM madel_
Complieaheo Paterm and oxtat iSSUe
daudapnent ANNwa Haeult an
vep licae t orkin.
unman Gradn T orKivqr ol ANN are_eckems
similax ase a [Link]) nchuobyks, ol
o Mut idetical Ls iadhcle_facses olel
Noranhaayks
Detaon
daNualAahwoyk CANNs)are computi
SaenasAmpice a GiolgtealA
6ins
ANN sbased on Collechn_af_tonned
Lu ehecdes (alled aHilkctal neutuoy whid,
lesey maelel hu nwun na_biolagical t.
Neuron ane alyes pitally hewe a uerýt Has
adiustaas leuning aceeolt tL wat latea
ar deccases hength_ol Signala
Cennecticm

Sie logica Nenay 8


nerve cell neutnisa speaia Giolaaieas ell
Hat pocek elnormakn lccorclng t an ehmahe
aes are huge numbez o/ nukosuppY0ximukly
10"wH umennusinhrconneehiMsa2preimaky 1no

larki
ulorkiugof a iSialogîced Neuuona
ple newuun conaisBs al A talloing FOtut
e an explu
urKi
Denosfhs re taee -Ifke anches apenaibk.
tor ueceiving
conneko to n Hax Sense e can ga

[Link] A s Hcell Gaa t Munun


s ponsible r Proceuin oino
torom enolies.
3 Ake a cable
neuron SAnd winrtmatus

1na2 se is H onnechion Gekueen Haa aon


and neurens dendiAA.

Loaical lomputeukiens h Menea

T Gtckal neuLun s+mply ahak1 toudpt


when mere hen eain numhty o h
inguh are ative [Link] and PR
Ps_shoused

H een Such asinpikied medel _ is


pessit le C a nekcvkoarihda neus
ompu gt ppssThm tau waut

Neurans Canntehim A AND


VV 0R
ci NOTC

A A
C AAB CAY B C-AAB_

i lelislekis Guflel a le NNa dad perorm'atica


La mputionsagumin - dat N n
acivael when_tlesast npui
xe ove.

k s t netwok on let simp l a


idenikanhon RA newon A sacfva kdHn
Datc L 'age no:LI

neLn C aeh ackiva [Link]


aswellsince ve,
npw 3igns xumnuon E)
neon S ol'

o Secamd
Secandnakaoxk erkoms Oaica ND:
nLon C i sackvatol_on nLatk nu
Hond B are acvaeLa sinlenp sighelJ
is not enough achvat naon ) .
"

LME koovk
*

achiva kol ehy wuon


B sachivak d_lox bsh).
ina iwe 3uppose Hat an input tonneehb
CLn nLnon actv whie
Gblogica hlwzena,Hen a Laund hutek
t
_tempnbs Aighty mere cmplex lostapepsi
menA sahie
and
all
a c h v e wkn nwm a P anol vite venka

Th [Link]

T atins leaening anc A-kkctal Inkliigence,


h o nis d uld: lackel au AtiAial Neus
algasitem Rnables/uhanu o leaus
elemenhaselprees Ham u G ant aluning-
[Link] aon ln laclualei
dicus in-oley)h. hnowleb preepkun-

You might also like