You are on page 1of 5

Assignment 11 Paqe No

Date
Drbine sup
Explain the issues coding classibication and psediction
Ond conbidence in Associalion Rule lining. Hou oe osSix iation Le
mined bom loxje clalabases

he bollowing pacpiocessing sleps ma be applied to he dada to help


the clacsibica hoh D
mpaove the accuiacy, ebbiciency, and scalability ob
lpaecdiction proceSs.
(0DataCleaning = } zcbers to the pacproressing_ob dala in oTde to move o
educo noise and Hhe thealment ob mis.sing valus. Although most clasibicaion
data this
algoithms have Some mechanisms bo handling_noisy o1 missing
Step canhelp aeduce Conbuse dunig leahing.
)Relevana analugis -lany ob the_athibwles in the data ray be rdurd ant.
Hebe, clevance_analysis in the bnim ob conelaion analysis ond attiibute
subst selection,can be_Used to delect athibuBes that do not_Contibut to the
clasiacation_01_predicion tasks
Data Tansboamagtion_and Reductionhe data may be transbormed by noamali
zaBionpanticedanlywhenneual networks o methads inyobing disBance
mtosuiements ae usedin the leoning step. DataCanalsu bereduted by
appling_ hnany othe methods,1anging_bom savelet Ayansboamarion ond
puincple components analysis to discietizahion techniques, such as binning,
|histogzam onalysis and clusieing
Suppot in Association Rule MiningT}is on indikation ob houa brcquently
Hhe items appea in the data.
Conbidence in ARM -H indicaBes the numbe ob timts the ib-thtn statemers
07c und true.
No days due o iapid qrowth oh data in onganizahions ,ex-tensive cata
pIOSSIngis a centralpoint ob Tnboimation 1echnology. ining od Asociahion
Page No.
Date

Tules in laqe database is he challanginq task.AnAprioni algonithm


1S_widely used to bind oud the brtguent itan seis bIDmdortabast5-.

Leanin Uutcome l-1 leazned about issueS Tecading classihcaionk


preciction also Supp01t ko conbidence in ARM and hou or ARM mines
biom age databases

Explain the Naive Bayes Classikication decision tuee induchion ond Árihe
-ial leuaal Nethaok olganithmk-neaztst neighbour classibicarion and
atnehc algoithm widh opmopaiate éxomples Disuss the adyoniaqes ol
al these_appzoaches

Naive Bae's Classiicai.on14 is o classikhica on technigut based oh Baye


heoem with an asSumption o independence amanq piedichms In simple
Hermsa Naive Bayts classiia assumes thoai ha_pzna ob a pazticvlar
htahue in a class in unzlaBd tn the prtsn.ce ob ony othe btatze
o examplea but may be censideaed to bt _an apple ik it s rd.icundand
|aboud3inches in diamekr. Even ik these beatats deptndon each othez or
pon the exislence ob the othu btatursall ok these properties independen
Contibute d theprobabilit that this but is_an apple and hat is whu it
isknouon as "Naive
Advontages
TH is easy L bost prdict class ok kst data set1t also peabom wrl
in_mult class paedichon
LWhen assumption o independence holds. its classiia penbams beta
Lcompae to othu models lk lnsishic Ttqrssion_and yov nted less
haining data
perbam wel in_cast ab Cattgozical input vaiables compazed fo numeain
Page No.

Date
assomed.
Vaxiable is) or numeaical vayiables.no,mal distsibution is

Decision hee Tnduction- 11is a supexvised leamning methodUSed in


data mining classibication k atgsession_methods.21 is atee thad helps
US in deision making_puxposes t coeates classilbication o 1ecession_moe
as.a teestuctunt11 sepaxale a dato set ino smallez subsetsand
at the decision tate is steadily dexcloped . The binaltue ith he decislon
inodes and leab nodes
Advontoges
does not heed scalingab inboxmation
issing volues in data also donot nbluenceHhe process ob builcling a
Choico re to any conside1able exBent
A decision tate model is outomatic and simple to explain tothe ttchnical
team as well as Stakeholdeys.
Compoit d o other algoithms,decisiontiees need less exeation boy data
piepayorhon doingPpiDcessing
A decision tee does not cquie a standaadization o data

Atibicial Aleural Netoonk Algoitn-1h also chened to as simply


a "Nleusa Network NN),could be o pocess_model suppoated by biological
neural nehuozks.Ttconsists ob an intezconnected colle.ction ob azibiciol
neusons. is a set ob conneced inpulloutpurt units where each _connechion
has a weight ossociaed wth it.Duzing the knowledg phase,the nehoouk
acquines b adjusting the wights to be able to pxcdict the corectclass
abel ob the input samplts
Advandages
Thise have the abilityJ to piovide the data to be processed in_pazollel, which
mcuns 4heu Can handle_mDIe dhon_one tosk at the Some ime
Page No.

Date
hese have been in 3Sistance, his means tha the loss o one01 mpi
Cels, Dx neuxal netwosks, imbluencosHhe pezboumance ob ANN.
NNs QeUsed tostoe inbuzmahion on-thenetwonk so thad,evcn in the
absence ob a data paiz, it does not
mean that -the netwonk is not
genexscin9 1ules
ANN& anc adually being_bxpken cloanwhich means that they willnat
Sudldenly stop Laouking_and these hetuozks are q1adually being brokan-
Wt me able to hain ANMNS hat these nehoosks leaun brom past events
ond make decisions

Kheaxest heighboin classibicahion - KNN is one ob the most ustd le01 ning
lgothmsdue to its Simplicity. KNN isa lazy-leazning. non-p01o.medyic
algorshm.H uses -he daBa with_soveial classes tn prtdict the classibictian
ob the neua Sample point11 is non.paomehic sinceit dorsnt make any
assump.tions oh Hhedata bring studhed,it,the madel is distibuBed bom
he dat
Advantages
uick Calculation time
Simple algozithm-to inttaprtt
Veisatile-vscbul bi tgitssion J classiicoion
-Hgh accUacy-you donot pecd -o_compane wilh betla-supcavistd lenanin
No assumptons qbout data-no heed to make additional assumphons
Hune seveial paomett7s01 build a model.Ihis makes iHcwcialin hon
linear data cose

Gnenthic AlgozithmLis a heunishie stazch_moclel usecdin arhbichl Intllig


tnce and computing sused bnsbindingoptimizedsoluiansfo ta seaach
poblems bosed on -he-thtoy oh natunl stlecion and exolutianay bialoy
Page No.
Date

Ihe Se at excellent b Staxching thuough loge and cormplex data stt s

Advantaqes
seonchhom a population ob poinisnot a single point
USe payobb,inbormation, not dexiatives..
11Supportsmulti-okjective_oplimization
s e pxobabilisti. onsition 7ules not deteaministic zults.

eay Ding Dutcames-1 leaTned about Naive Bayes classibicahon


decision dre inducdion ntibicial heuol netork,k-nemest nelghbouz
classibication Gencdic algorithms wilh their advantages.

3 Lxplain bllouwing teams in detai ls: Paxtioning, Hiey0zchicalDivisise


, Gnid Jo Model based clusBrinq algonithm wilh appopxiatk examples..
Discuss-he odvandlaqes ob i dhe appr0aches.

You might also like