_—_Almtgament fo thed ts Date Science ) Algo exploén olala jeience, fat, Dele agnce & the area of shady whieh ——iguales &xtryacting to.teg hts Baca eit arnt dota ustne voFinus ‘cient bic methods algorithms and! procenes, Th helps dou to discover hidelen’ patterns faom the” nox data The téym eto Screnee hosVemereed becauge a ine Boolation af toathemadteal “ptoticties ; a oe Tt & an Coterdiscipltnery ielal thot aflows you to extract fnowte de V Prom Maruchired oF Lunsteuctuxed data - There axe fome Atco that axe necesrory to ony (cutk{al 28 from the clay = [enclens FORSUNG | : tedet [Data cndosifond, mere me iL (Wedel evotahon Deck preperation |. 5 : [Data | =| Geplendlory |S \ eehaga (__ Dotto Anode CR —© ‘Doctor Collection - Are. fe Slalement the maen Jak data hak can help sto a frontprtalton, Some times clea ef ae ace ome ind gi : — Cleaneng Cost o ra usorlel_ dota G_not Afqudured neq usr Cleaning and eae ento Abructuree) data b8fane db con be wed for ony madeling * Exployaloxy Dota Anal ysis — Ths ch Fhe sre Ete tohetch awed bay the Patterns, éy the ‘data, Alio we try to canal yee _suffanen t foctoxs which offeck the, ( eat . “Hodel % Quild 2 Othlerent ML og 4 well os | tedhige «dl hey _| abel gous colen te Bo te Can Cas ten te{y Com ployed —__ in the deka ae well iat steclZom us at {ew ureny | > Model Deple cent Arex a trodel & developed we deploy if cand mo, clon cA paafommance,(Gate oar _ Explain a aang B Dato Analyses “fintt Exploratory Data Analysis (EDA) & a critical ata frock fee in the Mota onobsiu pro cert involves Jummarstziag and autuolizing te main Chore ctertsher and Pptte ana ha datoret “The primary goof of EDA & to gain a_bettey undarrtanding Y of fe dota identify potentta! frends Gnd onomalies and gen enevate ‘tnitahta that can quide uvthay Tonal th Hee ave the keg asnotes of 0A : 4. Dako Cunnmesézectton : EDA start wrth a Jarome of the dolaset'’s boric Jhetates which Wilude measures uch ay mean median Jtandasd devtotion minimum Maxtmunm and quartiles for Qumericaf yoriables , 2. Data GauslsaKo> ° tt & & [andamental Component of EDA, Vuualtzation Such a histoqarom bax Charts box plot _ kcattev plot neakrrases ave wed neprerent data qraphicolly 3. Tdentificalton of Vadtexns and Trend! : EDA aims tp tolentty palin anol frends within the dota The involves lookéng dow clusteyg ovnetedian _Aearonabtty and’ other Afructures —to the data "tha con —érfoarn Subsequent anabytiser: Sa 8 TEEN ia ws bd Check * Cand tateqaly — Se Siar ula Hag gully EDA helor asseaa “The of the ‘dala eninge Cyuey Auch od dipuicade yelordi fata 2a fret 2¥TOTA and en Consu fen uies Qe Explain (Cn 20% Reg ~eHcoa uouth geamele Pi Lineoy xeqnestion ts one of the aajvext ond pulor! ML algorithm Tteua dragedico method Wok 2 wredl qox red. Wwe analisri “Linear reqrexyton maka medickons Jor’ Continous [veal ov numeric yortable jut oj sales, falawy age. puoduck pwice, etc. + Lineay %egsegcoo oloowithm Lhows a tnoer welution-ship between a depend ont Cy) and one om mance inde nendent Cr) omoble hénce called 03 naar yoq ¥ion 4 iindependent variable (%) Hate crate Cables : + O,xehe uf = Te y= de péendenFvamekt-e Y= endlep end 2 Vomohte ab = inter cepket (ne at = non ainemion Coeficiente @_ = onda _exidox Exomple: Cuppose wre wont to predict c f t Dtudenk’s kam score on theiy Aumbey ef houy diedted The Jollevoing loteret ave : . Hound Studied Exam 5 Cove 2 So 5 Go 4 +oSe ge bo by ——Sten Q)Callediats Mean SS —Ste p(@ Col cudete 3lope Gy e os (3) SI wags Geel =-e8 Sa -3) = Yo-t =(bo0-60) = 2-E = Gi-3)-] $5=4= (te-8) = 10 b= Cheto) Coxe) (ix po} Cee teO aed ae . G.= jO+oOr0 = oe = lo (+0 +] a bep@) Callallake Gheveapt b, = y by xe bo = 60-[luxl) = so(Cate: Page: aa Novo we have be ond by wwe Can make ~precticky on — if Akuclent sfudias foy Fhouwreg (c= 4) we'can’ predict frovr exath Ico 7e =e B50 +o x4) = Jo Ul btuclent shudias A hous we predick ey Ream Cove appre x 90 . GA Whar & Daksa Wrong Leng proces? Explocn with the help of ex oMnpl Bs” Doct Wrecaitng. alio known pa data roungeng "FO brondto wring cae nizing seus data “enta a Ltauctwved ” Joamat Jub babte fox onalyris Te tavalues handlitla cau tag values dealing with oudtiors, formating Aska and " mexg ging dodeseks aenong _ otino-r toska UDota " Waangling pa Cxuctel step in doa pxcpavekton\ f d and & offen a time-Conkuming part tsp SHS, Jciance Yond analytics lehis” explore" the data Wrrangitag “pracers woth. 2x cmp les ci CR : aka Collec hon + Example +. Tmagene you ore working wth e-commerce dak You Collect hata fren Various ourvees ingluding customer ‘orders — Produck information and chitomer review. “The data thou be wa oki ff erent f formed kuch ay CSV Joon ox data hares 2. Randle moiiatng Dota - Example: You decide tw hondie the mixing vobue, in the “customey age” Column, You can Chore, to }ill_them with the mean og@ Ov une a more odyanced im putakion method 3. Vata Trent formakion ‘ Example + You megink need to convert date and time data tnto a consistent foymat extract Aperchic endorrmakton ov Create. new feotures 4. Daka Ex post p * Example : Once the daka clean and trams lovmad to Your joktsfac ting You Mey export Ur bo a naw file ox dokebbaye” ready fox analyrcssz What & asus oxbractton) Explata the ou eokure Oxtyockion, Ba? Feokunre. extractlon & a dimenstonalt veiton technique used in data a and achine (Corning, Tr envolyes transforine maa dato arte mony contotn a _ldvqe nuimbey P vosrtablex tale a Smalley set i informative ond tadependent features “he purpose of fective ox byberton % to Coptuve the Seren bead intormotton foromn the dota while veductnag of complexcht Here ave Some ken advantages of feliture. 2ubvracktun : 1. Dimensconaltty Reduction ? om Jelecking of creating a smolkey seb of features, fou Séenpli fy the dats visuolt2d and analy ee. 2. Improved Model Performance : ML models tecoined on a reduced set a features tond tobe move ietent _requive “le rejowy La Yesulting tn better predichive Ae ATOte 2. Reduced Computabtonal Complexity {_Eeoluce 2x trackioa can “Aignifican ty yeduce Computational reoures yequired Poo dota Ona lysts and modeling :om (Lc “ve duced coe flan Fo data when dealing Cacth jars S. Enhanced Voualizatton 1 Reduced - dimensions} 2natAcom, dota 4 corstey ty valize, to Creote meantag ful elo 50h One tiysca and S Chowty Hot (Convex fodeg be more. Hectvalyyy Ge plata Dectiion Eyes weth thehelp of xampie. Sappoie Theat a Candidate who hora ob ey and wont fo dectdle whether he thould accept the. offer ov Not (o to jolve thir problem the decdscon bree shart with ywot node Jalacry attribute. by ASM) The oot node Aptit decy the¥ into the fext olec&ton node (disteunce. Brom the fee) and one. leet nade, based on the. cofwwo speaking labels “The next deciton nade jaxlhew aety ue iacke one. destin dode Coab. Ey Te Gn one node Lnolly the “dlectkton no Sout ne et te see ncde(Cf\ccepted offers and_declined afer |} Constdler the below diegryosn Selovy d behoeen ys0000= troson \> Vo e 5 2 OM ee hear to Dectiruct offer Rome ye wo a 3 pectinad offer Vrovedes Cab tt faety WD Ww Accepted Hecicnad —Cfey Oftey22 What Ti Weaandata a : a See ae Sui O43 o A endatton Suit 2 for natlon | ——necomnnendery Agiant Hr qubelea of cnferranien ——filbertng systems the provide perdorn nS —“recomfnenflation to wrers. thare Aye pre widely “wed th verinuaonling ele om — Xp help uke dds cover Qroduds, Ts ha OF CBntent thot Hey might be interes Lared on Hnetr past behaviav & prrtferences Recommendation systems axe. “enenttal | fox Enbon cing Uper Ox pextences and Cncrearen engagement PAO commerxce , soctal media & mor 3 j d i i 4 : ] j i. Thsre one Severo? typy Pa Recommendation i J Sup tert j j 4 i 3 id j LL. Collabowabire Filtering fl + Uie7 bared @llabovative Filfering : Te Recommends crema bo a wrew that imilay Wer liked, 2 Trem bored Collaborative Fillering : Tt yecommend: tremy that ave jemilay fo the Oneey the user j hoy onkvacted uwethin the part. 4 = : q‘on tent boxed eliatge These_method aecommends Crem Sinrtlor to those the uger hoy Ahown: inderert tq bajed on the —_ pyivensy actton oar Qx ect feodbuct : 3. Hybatd_recommendakton Systems 2 Tt Combines Multiple yecommendation ee jp _ leverage the Abyengtny of both Collaborotu and content -baxed methods,“ 4, (atent- Pwoare Recommendation : These Aystemy conitder additona) contextuod tnfonmedtion when maling vecom mendations Céntextucl factor Con tnelude time, location, device and wiser behavioy Ww different: Contexts SE. Sexton — Bayed Rocommendation eee cic, Aystems focus on Providing WeCoMmeondeation within “a finale user Jepion hore the. user's perfogmance, may Chonge over time . P=re RE , & Explora d+eature, Jeleckon alt ee ee a pat A analyse where you C a_jubset af rags molt ere Levent feokur€s rom fhe. ‘oat inal jok of featuher, The Goal & to ¢ ae mo del Pezformoane v7] and ebhource the Seduce ovey tntoypaetabilth rE he model, Here's an overview of fas ture, Jeleckion algorrthons « 4, Filtee p methods: It eveluate the imperrtoree, of _f atures indlepenslently of fhe machine Tellentng model, They +Y¥pctolly envoWwe Atotistieat testa ov nrontting} boxed common hechnigue tacludles Cht- Square text Trdo nm atten Ghia CGrrelotton Coefficient: yamence Thxethaldsirg 7, Wrepperv Methods LE aysery fJeotusres t mponctebe by Comiderring they (moa ta the genformance of a “pect fic * machine ledening codé| “The ‘thos a Search Strategy to din the “optimod Jubset of footarts ; ViVi NN | es | |—Cemmen technique tncledes Foruserd Jelection “Rackwoord cleminatton Rocuysive uss Eiteinakton henetic edger Ham. Embedded Methods Tt Cnteqyate feature jelection within the model tratnind Proceta “They select featare durng modelY Constquction Vferkeng Vento acco®ht both feature tmpoatoncd and model peaformonce , VCommmaon Hoch gush énclude, Lake Rondom Foxext foatsre tmportoun ex Cnodéant Booting 4 deat +e impo rhonce, ty batd_ Hethod TE combine elements of filtex wropner, and 2mboedded method, “to select the ine fecurucres | Commmnan techrig ue tactude Rarvuty Figo cH en,ES Ge “Wha de you under land by O Kon? Also dite uss Aone Oakh vl tools. Deka Ves 5 data {o provide, dngig ne Pol ~attion of and tybads that moy not bem appayvent én row Goto Tt plays a, critical yolé tn data _ondly helps people encludinge es analysts daka Ascientkta ond cleciton maker understand Complex doty more of ficientiy and make tnfoemed de cktons . Daka Visualization Frons{ own dota ents, Vehucl olementa Ieke Chota, cavophs , (opr ond dorhboowds, g Some Popular doka yeuolPoution tools trcbude ie izakton ts he ga phicel sepprciatn J. ablean: toh o powealul tool wtth wgerm Peyiendly interface , TE ohlo wl Users Tel crent® inbvacitve and Ahorrable daah boards Charts anda Qro.ph + 7, Powey BL; Tbkta buxtnas cntellcqence Micro soft: tol that provide entrackive L ond cultomizable data Ysuoli zations |Whak & Hochtne Tag Dicxp len See ee enaleg 7 Bp achine leaving % a jubfiold of AL thak casts onthe development! of _algavethmu Atatesticod medéfs that brabte Compa uke ra to _[tacn male precdichons orcdectstons wetthout being ox plict tly Prog vonrned | “Th pvdey wWo%d4 — ML YX about Cweoting Austems ean curtomokt 4 PApxe thoiy performance pn a tuk Hievo wah 2x HOGAN, and dota There ove Several types of ML including 1. Superveseo} leontog ¢ The algoatthm (4 Fyatned on a labeled dotaset Which Vmeana gach enpub data potat th ayoctated weth a covreroonde ukout ov target, The goal & fo [2a%n mapping fx0m inpy to oukputs Common opp Cotiahs 4 ane TImog 2 Clemificah ion Apom arias deteckion , indech aecogni ton and alge on toskste Te deat wotth te unlobel data where Algorithm totes to edentily Matte «nr Akaructuses or! qeletfonakins wethin the clots witho ck Apecific guidance i the foam of Corget abel, iY : Common appltcations Cluatertng dimenswadih, greductton , Anomaly detectton asd fo pre Modeling . eau : 3. Seme - Suneryered Feorntng A TE Sa Combination of both Auper vexed and Nu peye sed Lecvrning Connon application ; Sentiment analysis ond cevtarn Cases on which label data tA hearer, A. Deep [earcneng + TE 4 oa fubset of rit Thot whilizer otek Qewral networks to model and Jolve- i complex Coak { Common aychitectes: Convolukonsl Neural Networks (Cnn) fore Crnage onabyyis . =Tat Deveck Discovery of communttrey en a 9.20 ph abe kno wn {8 Corman. poclectiva ov clustering to network ang! the ——pyeceas of tden Fifer ano oy Cormmuntttes node, within a “qrtkph | Where nodes are wow densely Connettesd to ach othey than fo Medes outrede the Community . Treye. Comnenundity often ep Cluste~4 oe {Se chtonod mood 2y cn ae ran and they prevede yoluabte, tartg Wb, tqto tho Sbeucute There ive arto ug approaches to deh cower Coenrruntti on o re Ph . : ae Hodulomty = Rarxed Methad, 3 Modularthy ha (gasure| Hee quantifies the, _quaditty Sef Corn unt: drachure ona gaaph rdular? bored method, alm to Moxttrize the, meajure 2. 5 pectrrol Clustering : It &a f2Chnigue frat leverage, the i960 voImen ond LXgon vectors a Qyaph motyvx +o porhton the atop indo Communi Eee ,Qe 12 How te ne preient Jocral Media 1.5 2tphs fle: ] So Rap Nefworks con be jcienlly and — LYjectively ve prcientakten os _qrophd Where Tend’ viduals oy 2nktttes aoe orspocaced GA Nodes ano) thaty connection av. cater ack toy AVE yepresented as edge between tho “a The Graph bored ceprexentutton allows | for [ho ahd Vitaltoabion. and modelia of Aocfod ae fotion Anions and Atrrwcku veh! q Heres & how to represent 3orctou nefworksd * Or avcuoha : 1. Nodes (Verkees) + 4 i i Each tndeviduol ov ankty tn the Joctol q nehoork we pacented “as a node tn the) q-ro-ph 4 Connchony oy urbeyconnecttony blw endividuals Ore vepvesented ay edges tn the qneph.. i Edge Clink ov Tu ) 1 :Tt 4 an exam ple of Eng focred neFooark ~The entittes ave tle nafles through G The _aelatienaktp whith we might Yaink of as frends yepresent by edgey BH fatends weth Ai Cc and D i ©) bom te a Showh note on (a)_‘Vatnctple + Component ana \yais Patnceple Component Anatysy tha démensionabh yeductton technique Conimonly wed én ¢ data analasta ane Machine Jeavning . Ze oto “to yeduce the dimensianodit of data et while preserutnag a Such ef the _data’s ont. g nod Va rtal porxuble, PCA achieves thy buy 2 Uiaing {ne principal component, Trak Cap furd! the moat S24 ne ecant- Aource of vomakton tn the data. | cA &y used Joy lata COMP «e4A2on + NOK. er ckloq and VUuali zakie, Os well Aime u (ying Como lex _clotayeee Gena aguas Wel ular Vol ae eer na . echartgue that de cmpose & Mate —Hrxeo"oter marta. ee ce pd = actor n ng of ‘he ont c aD WA cA Svh fon dimen bron oly eduction Uelection and matrix approxtetokton—. TES poate curlagly helpeng ta seducing I the dimen rio Aol CO of dota while stlorntng &m POX potter SsvD Applied TA veriows fielda, tmoge pre cetirg? Nokuval languorg2 xO caning, Joand OCo mamMeondakion AyA tem : ) Dmensieraltiy eduction t Te ho qeneval term sf eee wed ts) xequce the ho. fob Ox yartable tn a a _much -velevont th fexrnalion he, pr mory coal ey to Larges clades 4 odleniatbe, dato start ah totebe “EXP sats oes fovest wth he sakes Cramp le. Suppose YO W0%K ta q healthcare domain and You’ tesk A bp tulldia model that predicts whether a podtedt haya woorticutay cryeve beyed on Fray \medtcot nis tavery fep 4d: Daka Prt Aaa ction Age ender Symptom, Qeuulé Ducaye Outcome xs Male igh Qo pothive. 32 Female low Wo Neg ative Sg Mole : egy \4o wose ve Sbp=2: Qutidfng a Random forest 1. Data Aplctttng i_spltr the dataset Pato buoe pants Y broining set ® aket set The tryar cing ver ty wet to build tre2: Random Foret Emsemble A Readom 1 Foret Con,yts poultCple ie Aecty ton “Freak, Fach byvee, ta _trrotned - On Jan dam jubtet _—— anol the fectures Ca fe betcly ~— 3. Decksion reas: Cach deciiton bree o bull to predict the disease puttame bayer ona A wes, The vondomnaw tn the and ine faotuxes pyevent overfitting Li oe Mechantine ! To predict the prediction the Random Poryest combines the Oukpoca a all Srdiutdueld decthton kegs | s. PredtcKon > Each dectiton ee ta the torrente provides ttA owedliction and the Moa om ty Vote, determine the, feoot outcome!“Step. he Model Evalnakion a 7 We Cvetpote the Random Foveik? peslormance on the text Jet by Comparing ty prediction to the _actrust AE coe outtomes. 2 Explain k= means algowtam : = means Ba Populay Cc ustextng_ alaoxtthm used tA undupery ced mMachiney leaning . “Ltre Qatma QuBpose A to pore tition a Cakarer Yinto tine Aon~ over lopptng elustena Where: data points uoerhen the hame Cluiter are mone Aimilay- to Lodn othey thon to those ino toy cluster The clgowithm t wédely wired dor yor vA Op plications tn chasing Vourtomay seamontake Cmage_ COmprek4LO Cand anomaly deteckion How the ke mean alge atthm werxks + st¢o te Trtikali2abion = Cheese the Number of Cluxfey Chk) = Tatitaltee Cluster “contert,_ Step 2 1 Amcgame = Dian data patais im West Chad ~——S#2937 Updak = —~ : a ee 5. ~ Updeke “Cluster centre + 4 i Step 4: Con vergence, Check : i = renesk Jen 2% 9 until one of the Ate eping critemia met + o The Cluj ter waregqnment no long ex, e@ + The chytey centey no longev more erqnidy ; + A maximum numbey ef tkervation U wemcha Step S2 Out put ~_ Finol Clea tereing 2—_ Daka Actence Ch a mubttdieatnory Aield thak uses Acten Eefic_ methods, aladaiPhons Qywocemers and Austems fo exhyact knowledge ond Cnata bot tom Atyuctured ond und tou clurced UTE encomp Hea a wede wonge of actturtie, thcluding data collect'ton, ‘date cleaning dlutd OSOU44C%, and the develoament sf pre dictive _ model , Data Ackence Envolvy exfaackion casight fy errevof Q Conc 2m; fis Pat vere ; Collecting ond uitag pextonol . ene: Enduring individuals pees ee Chota al b2 wed. 7 5 me mig neh 2 bre Vente onthe __ -Ries_ond a OY, Lin Cebench and we “ucpeuthor&rod a — dd yeitin Ming — Those, 2lh@col tyuer & Cry OO to onAune %eAD onsible, ond rvs ati data Acience pxocksees {2 Discus about privacy jecumty and 7 2thécs vocth rep eck te Data Adan ec - ~ o / / ss Potvacy fecurtty ond etnies one, Cyuctol ~ aspecty let Acience a ae peafadn toy the Collection , 9 K leks explurt nae aipeckne to _cletosl s Ae Raw oc, : i ae g 4 3 Part acy Concerns revelve around Til pro tet en of todividurol’s peronod lormekyon “ Haka Actence of tan eat oles the collecttCon and fxd cee of Zengetere chaa toga OK Tlen bifiabre. Laformotion 1) 4A jeamthy 2 . Def Accu] oy about Aofeguors ding de thy wnarkhax unauthorized © geen reaches, ond data thet — TE closely (tanked to Pm voces Aecuty batach can CO mp0 oeile. indivittel, personal information , Sees Ethical congideyatton ta dota Jetence, Lacon a _twtde yonge of tuey tack fatanepm trans daren acquptakity and tro neigonitie Tae The th Fewsectibn Ee Aecuraty ond Otter Auonce A ont] ler burld & fa eee protecting ead vided and 2nhurngY the TAP OMAT DLE awral latotal we af YateSree a Tes Ei aE ob aed cen Sao Took hack of Data Jc defence —=y ° Emergence of —b&4_clatka Cloud ae and © aduounced Ymmcichtne [eos 09 3 ° Tater digero lina rs apprech woetth diverie applt caCtcon £_ Qyouwtng empha it on ethical Conc ms — and a@rponscble AT i ‘ Improved cota [rtevocy among Oreferionals g qd Collabovelion belween data SRG ond domoun 2xPevb A <5 Next Gendvobton Poko Acton eae o lSvesoqing even mone advanced toch, g . S troag ev eMphakA on twa and etuel: LOO WENG « 4 - et yaoonstbie AT and 7 works .(xe: a : cf te Ahart note on Partition ing a goraph, Rar tition tag Trophs ts a fundamental concept tr Sarvopr Fheow, and E Computey scence, Tt enydivey olividing a qyeeh ento olds piat Jubset, ov. Pagtitions wWhil€ pyereruu certoen ot pio pero Character 4bee, the cacigénal Qreph, This qroc application CA Yordous ieldy including petuerk Aeséan data Clustering and porxalle| Computing 4 ‘ ‘ Key poutk, are 4. Objeckre ! Partitioning to cat bube, we nodes ov vertex of Arvapsin kato diferent Jubsela pf Q Graph ky pe 4 Greosph fy pe dle pendd 09 pw blem lupe ef ggaph fo Clusling Yadiveet " dine Weignted q-vaph 8. Blaowmthm:. Vemerouw alg on thm Hoy bean _dewelo ped to: Jolwe Graph partition problem- A OSS DAD A — ————— ; Be ] Exe in Clu Feving of aaaeh ond Neig h beahoad eae edatios en Graphs pe Clustering en - h_and and paighbovisad Paxopertcoe axe. fe one “Concept | in q’voph theowy Seem They desenbe how nodes ow vertices &n Qa aro.ph fends to be tnrterconnected and j howd tre Atructuve of these Connections | Can reveal Umportant\ Mndinnk abo uw j fre Geraph topology and “relation ships. | a. Qua tering tn Crvaphs $ Cluyfexving tn ar h_aefews to the fondenca S of olen to 8 SDoR TE BS tnkeyronnedled groups ov clusteat Clustecr.ag iA_often ee us Eng Chua & mete, Wee} ud Leg Coet{i cid + pS Necgnbovhood Poopewtin of faaph: The nerghbor hood pode ino 9 Yph acne fo te. The Atudu of : ‘ i—ne 4 helps undated — = Vv fora rq hborhoed Pepe Common a erqin boxrhood Pre perhos eneladas? ei crc he local shyructure — deqnee Cotamon Nerqgh bayhood - OH&dtance = lo cal Cus fernang coefeant: