Chapter 18: Data Analysis and Mining

صفحه 1:
Ohkopter 00: Ona @adeie cad Dictary 

صفحه 2:
Ohkapter 06: Oata )19 ‏جه جاع براه‎ Dietary Orvisica Guppont Gpstews Orta Boripsts cord OLOP Orta Oorekrustery cts Dreier Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wo ©Sbervehnts, Cork ced Cnakershe 

صفحه 3:
Qeveiva Gupport Gysews BH ‏تما لت تال مه با و‎ dere, often based ot ‏و سس سین لیس ربا سوت مس‎ eee, ۲ ‏تس وا سس‎ © Dhol tees otk? ۶ ‏اه نا مج سس با‎ ۶ Do whew ‏تساه ام و‎ 2 Bere ire eB da ved Bre somber ‏سس‎ ‎۰ ‏لجسي سییر با یب‎ ۰ Geet (ee eerie) Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wo ©Sbervehnts, Cork ced Cnakershe 

صفحه 4:
+ Osvisewe-Ouppori Oysiews: Overview © Oats ander tobe oe stoplPied by spevidized took ond GOL exter © ‏اه مس‎ ۱ ۳ ‏ی‎ produ category oad ‏رح موی‎ whol were the totd sales in the fest quarter ood how ‏نت وی وا ول‎ the secre quarter host pear © Oe ‏بو تساه‎ ewk prixket ouewery unl puck einer oa ery Bl Grteted omnes packenes (rcp, | G++) ma be terPared wih dacboses ۱ Piekd, but cet covered kere ۲ ‏مت نو‎ seeks to discover hoowiedye cuiccwvaticdly fa the Para oF statistical rues wed puters Prow horye ‏لول‎ BO dete werchouse uchives iPorwativa quhered Prow wulipe scurves, ocd stores header o UaPted srkeuwa, ofc sine site. © Seoporteat Por horge bustuesses thot yecerute data Proow wulipke: diisicos, possibly ot cattigke sites © Dota way osy be purchosed extercrahy Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO we ©Sbervehnts, Cork ced Cnakershe 

صفحه 5:
+ Octa Baddow rd OLOP Bl Ocke Busbard Provesstay (LBP) © ‏وبسح‎ eerie oP det, cewice chic iy be svercrartzed ord viewed it dbRerect wove ‏سارت لس مات مه‎ deka) © ete tot coo be wodeled us ‏واه حول‎ oad ‏تا مت‎ are ‏مت لین لصا‎ + * Orwur wives ۱ ‏ی‎ suwe ihe ١ pes be operated ‏مب‎ ee. the uinbuie avaber of the subs rekon ۶ ‏ای م0‎ سوه ج) جاتو سس ات من ول با بقل ۱ ‎tere) oe viewed‏ ١ ‏دج‎ the itbutes tew_onre, ool, ocd size oP he sake rektic Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wo ©Sbervehnts, Cork ced Cnakershe 

صفحه 6:
‎oP sules by iteco-cawe aad‏ مسولیك یمن او ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‎ ‏و ‎Dhe tobe cbove ib on exnople oP a oross-kihion (vroe-tab), dbo rePerred tow‏ ل سوم ‎the row headers‏ توا اه مرو با ‎Ocke Por oe of‏ ‎Osttes Por carter devecsica uttibute Porc the ooh headers 6 ‏من لاه ات موف‎ ‎]9 ‏حلاصت امس لهك ذا دص‎ one (aerexies oP) the vokes oP the ‏ما اه مس‎ speoPy the vel Ez ‎Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wo ©Sbervehnts, Cork ced Cnakershe ‎ ‎ ‎ ‎ ‎ ‎ ‎ 

صفحه 7:
number 8 35 10 53 20 10 color dark pastel white all dark pastel white all dark pastel white all dark pastel white all dark pastel white item-name skirt skirt skirt skirt dress dress dress dress shirt shirt shirt shirt pant pant pant pant all all all all Bl Cross-bs oom be represeded ‏جه‎ ‏ار‎ ‎© Oe we te ‏سل‎ dis werd represen © Tre GQL:0999 stocdard artudly, ‎dues to place oP di despite‏ أن عدص ‎oP size ih regudar oul vues‏ ‎Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‎ ‎ ‎ 

صفحه 8:
و0 + ۱ © Ouakae o deem; we vow ٩ ‏ساسا‎ ‎1 ‏صمت ولد وير‎ be word we views oa a chia abe 28 48 42) small 77 ‘medium 53 | 35 | 49 | 27 | 164 large ‏و‎ skirt dress. shirts pant ‏لله‎ ‎item name ‎wo ©Sbervchnts, Cork ced Cnakershes‏ ۷00,06 بط 9 - مومسم 6 ی ‎ ‎ ‎ ‎ ‎ 

صفحه 9:
۲ ‏یوم و و لح سول با موه پم‎ ts cohed © Gheteg! oreutoy 0 orvss-tob Por Fixed voles ‏رای‎ © Gowetves riled ‏روط‎ porimked) wheo voles Por ‏سود سول‎ ore Pred. ۲ Roky: woviey row Peer-grocuoniy data ‏وی و صا‎ yrocukiriy © Ord dws: Che opposite operaio - thot oP ‏ما لوصو مر وی‎ fo ‏ول راوس‎ ‎wo ©Sbervehnts, Cork ced Cnakershe‏ 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 10:
مجه يان + © ‏مرا‎ on dkvewios utributes: lets dveusioes to be viewed of dPPercct levels oP detail بجفمل ,رو و عمط وا وروی وا ‎used‏ با مت توا ‎Date‏ مس با روج * هجوت ای ناه ‎clay‏ Region Day of week Country Hour of day Date State DateTime City a) Time Hierarchy b) Location Hierarchy Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 16.00 ©Sbervehnts, Cork ced Cnakershe 

صفحه 11:
Cross ‏له‎ )0 Wierachy total © Oress-tube von be eusiy extecded ty deol wily hierarchies | Co del dowe or rollup ooo hierarchy dark 8 20 28 20 34 62 5-55 item-name skirt dress subtotal pants shirt subtotal category womenswear menswear total 4 Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 

صفحه 12:
ههام1 ۵0)ران) 19 Dhe ‏ورد 006 وه‎ werd walithrewinnd ares ‏و وا رم‎ dota ‏ای بت با واه سوه لجی ,لت‎ OLOP (DOL@P) systews. ۱ ‏سس (60۵0) هر‎ 7 ۱ ‎(LOL)‏ 0را0 تیاب اطه هط ماس د دز سميج جات ‎can!‏ ‎sees.‏ Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 16.00 ©Sbervehnts, Cork ced Cnakershe 

صفحه 13:
+ OVOP ‘kopeweadion (Ovw.) B Coty OLOP ‏مومس موه‎ all possible oqgreyaies ta arder ty provide valor resp © G pace onl Roe requires Por doin sv oot be very hits ١ 6۳ ‏مج سل‎ by لمح مس اه هلوت موی وی رم وا ولو © وروی امس ان جم وو )مه سوه مت تم ‎cobr)‏ )مه سوه و ‎Coa‏ > ‎cobr, size)‏ بجو ~ or dl but a Pew “oederowposuble” oer eyaies suck us wedi - ‏جا‎ cheaper thoa ‏مج‎ ft Pro serail © Geverd opiketzaiogs avoluble Por cecoputeny culiple ogre gates © ‏ون‎ prep uie ocgrecnie oo (ewan, rel) Proc ‏وه و‎ ot (ieu-anre, cob, siz) © Con co wpuie agrequies va (jiew-cnre, ober, size), ‏برع‎ ‎oP the base skis Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 16.00 ©Sbervehnts, Cork ced Cnakershe 

صفحه 14:
6 ۵ لس + ۲ ۱ ‏اه موی وه مه و‎ rou by's oo every subset of he spectied ‏اه‎ ‎7 ‏(صم )مه مه له روموت موی‎ ‏تست میم‎ eng by ‏)له‎ cobr, size) “Dh cox tes the vara of eft di Rerect rman of the sabe retin! { (iewe-nnre, cobr, size), (tener, cbr), ) size), (cobr, stzr), ) ‏,سمس‎ (cbr), (riz), QO} ukere () dewies oa ey group by bet. Bl Por ewk yoke, the ‏صلب انج صا فيه الحم‎ Por otrbutes oot ‏بوسحم جا جا سسب‎ ‎©Sbervehnts, Cork ced Cnakershe‏ 16.00 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 15:
+ @rteaded Bqgrequios (Oow.) BB Rekitoud represeciica oP preset thot we saw porter, but was cule pce oP dl, oo be computed by ‎oped on an cate‏ جح حت موی مخ با ‎voke represeoteg ol, gad retures (D i dl her pases.‏ للم و ‎Rete CP he vole‏ © ‎poupha 127) wr stze-Pkry, Prow subs wpa by nube(tewonre, rob, star) © Coase he Pucetva devode() ‏ا‎ he sebet chur ‏توا نا‎ sunk walls by ‏صلم ه‎ suck ow © Coy. replace few-oe ta First query by ‏(ست- موس الط‎ 6, ‘ol’, secre) ‏سا0 لح 0 لا سواه 1 موم 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 16:
@xteaded Oqgequioa (Ovd.) BV. eh ‏وه مت و ی‎ every prefix of speviied bet oF ‏اس‎ ‎B Gx. العامة امعد سج اه وت سوه ‎Brow subs‏ ‎cob, size)‏ برط جحو مج و اه منت حطس [() ,سس له ,تست( ند میس ] 0 ‎hierarchy.‏ ‎caery) dues be ooteceny of ack‏ سروس ‎BG, sxppoee tbe‏ ‎tow. Thea‏ ‎axa ctor)‏ ا 0 ‎beaker bern‏ = ۳[ ‎pow by eke (cari, ter~ornre)‏ مه روا لو سوت ‎a Kerorchiod sucmmary by‏ وف اون ‎©Sbervehnts, Cork ced Cnakershe‏ 16.00 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 17:
+ @rteaded Bqgrequios (Oow.) وصجات بجا ضحمب ‎side‏ د وا جوف ‎cua be‏ اه لت وان اب( ‎vet of row by bets, orves procket of sete ver overdl vet oP‏ ی ‎Bock‏ © عبط نوا مج هم ‎velo ewe, oobr, size, oeo(aurber)‏ ‎Brow sake‏ ‎ow by rohip(iew-cnre), rohup(oobr, viz)‏ وم سا وس ‎X {(evbor, stze), (ober), ()}‏ })( مها ‎roby, size), (tew-nnre, cob), (tew~cnnre),‏ صصص ) ) - ‎(cbr, size), (ovr), ()}‏ Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wae ©Sbervehnts, Cork ced Cnakershe 

صفحه 18:
تج تمه بو له وه موی وا سل با رت ۲ ‎of pork ote,‏ وت بط لا (عی لاله مات مرا ‎٩‏ ۳ stickorid, rack ) puer (order by carte den) we oro ‏ویلبد موم(‎ © centro order by clus is ceeded i yet thew tn sorted order ‏تمه مساو‎ rads ( ) per (order by worker dewey) er sro ‏او یله مب(‎ order by s-rodhk ۲ Rochieg way ewe yups! e.g. PC stdeus hove the save top work, bots ‏که هو‎ od the ‏ص عادص فص‎ © © ewe ‏لو‎ does ot we cpp, 9 wet ‏سا وس‎ Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 16.00 ©Sbervehnts, Cork ced Cnakershe 

صفحه 19:
Radkiaq (Ovd.) BE Rocking cou be doce within portion oP he dats. “Cia he rents of ‏ليخد‎ thie eck seviiva.” web sida, socio, racks ( ) pver (partion by serie order by works deo) ‏ی وه‎ ‏وطاع اهاط مج‎ where sidebars. sida! = stideutsertiza.stickubtd ‏میج شید با لو‎ ‏و سل( ها‎ chases ou poour ina sti selert ohne BE Rachie iy doer oir opp ‏دمصي عسات بها وححه‎ Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏موم‎ 

صفحه 20:
+ Radios (Ort) ۲ Ober rechten Ricci! © perved_ rank (wihis partis, P partiivciy ts doc) ۶ ‏سطلمم) بط وی‎ dotrbutins) ۱ wil: prevediery voles ۶ ‏)ام بر‎ it preseure oP duphoctes) ۲ ©QL11I999 perwis the wer ty speck y mule Prot or call kot اسلا ماو اج وه ‎by works deer ule boot)‏ طس) وه ( ) مر 0 يمتنا سا0 لح 0 لا سواه 1 همومه 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 21:
(مسم) مسج + و ‎the Posting ctk(q) tohes the tuples‏ رم ‎he‏ و مه مرف و لت ل اه موه طب میاه ند ما حارط لوط اتمه با دا مور ‎OP tuples.‏ نوع هم ‎mno(sukry)‏ سا سیر ‎Prow (‏ ‎vfle(S) per (order by sutey) ver forte‏ رمك - ‎welt‏ ‏+ جه ‎Prow euphyer)‏ جاده برط جيجه ‎©Sbervehnts, Cork ced Cnakershe‏ 16.00 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 22:
Otedowiey © Osed to sooth out rendow ‏او‎ B Gq: woes average: “Gived sues voles Por euch date, colultte Por purl cate the wenn OF the odes vo thot doy, the previous doy, ocd the cent day” ۴ Oknbw epevPraios OGL: © Ctearehtics sabe (dae, he) ‎sua (vake) pver‏ رح ماود ‎(order by date betters rows (| prevedrg cod (I Polowtcry)‏ ‎Prow subs‏ ‎۲ ‏ساس سمرت و اه ام‎ © betwee ‏اه لت پم ای من‎ © rows honed proved ۶ rene betwee 1 precede od cured row ‎» Obs wih uckes betwera cured row vos 00 to mares che © ‏لس مج‎ 10 day precedes ۱ 0 ‏صم مه اه‎ ‎ ‏سا0 لح 0 لا سواه 1 66م 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 23:
+ Oratory (Ovct.) B® Ooo de uted wie ‏وم لات‎ BG, Crea a rekion neice (arcou-onrber, daeckoe, vk), where inher fe postive Por ‏سره له امس و‎ Por a wid "لمجوجه جما مم جماصتحص موی له موه موی ‎oP‏ ولا لت لب" و لصو همست بای رف سب ] ‏تانج" امجععه‎ peer by dhe rows \cboented preven) ‏لا و‎ Pre renin order by cermuctouniber, chaitete Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏سا0 لح 0 لا سواه 1 موم‎ 

صفحه 24:
Oda Dershouwiny © Oct ‏بسن رل موی بان بت اه موه‎ historicd data ۲ ‏ون‎ devisiva wobtoy requires ‏ال و‎ view of ol ‏بل امش‎ fockadkary historicd data BO daa workoee & oreprstory (achive) of Poneaion qahered Brow ‏و ون ,ماه ای و یی له موه لت‎ ote ۱ perwiis sity oP historicd treads: © GhPs deveivg support query bod away Prow trowsurtion provessicr, ‏مدرد‎ ‎woe ©Sbervehnts, Cork ced Cnakershe‏ 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 25:
query and analysis tools data warehouse data source n Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 

صفحه 26:
4 ‏مس‎ ere © Qheo ocd how ‏سمل من سا‎ © Gowree drives achievture: data sources trrswil ww iPprarciza i worchawe, ether ovaicwwsl or pericdirdly (e.g. of cit!) © Orsitcatcn drives orvhiierture! warehouse periodicdly ‏سوه همه‎ iPorwativa Pow dota sources © ‏جامد )ا‎ warehouse exuny spochrodzed wil data sources (e.x. usiery fue a as ane * Dsarily O(C ty hove ‏مومت اه ول عول ای باه‎ > Dardlurdaes ofr permtcdy do udieaded Porc cole trout process (OLD) systecre. Bo Oke chew wr © Gckews tteqratioa Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏سا0 لح 0 لا سواه 1 وهم‎ 

صفحه 27:
+ Qore Owstwowe Orskn wours © Qa cowie © Cy, cored wetches in uddresses (wespelns, 2p corde errors) © Derg address bets ‏مه حول بو‎ aad purye duplicaies © haw © propane yrkies © Darchase schewa way be 9 (adertized) view of ‏و موه‎ Bl Oho cht ‏سوه‎ ‎© Raw dota ay be to kane i store ork ۶ ‏له مه (ط سل جاطه) سل سوب‎ ۳ by query optieizer to use ‏فص ری تا‎ سا0 لح 0 لا سواه 1 جوصه 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 28:
+ a © Okvewivd votes oe vy ‏هه وه ری لطس‎ oad wopped to Pull okies ‏اه بو‎ © Resuhodt schews is coled u ste schewa © Ove cropicded schews structures * GoowPoke schewa! wuligle levels oP dicoeasica tables * Opestehaioa: culige Part tables Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏سا0 لح 0 لا سواه 1 موم‎ 

صفحه 29:
tore store-id item-id city iteniname sales state color item-id country ‘ae store-id category customer-id 1 date customer item-info 08 customer-id date-info price ame street date month quarter year state zipcode country Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏سا0 لح 0 لا سواه 1 وهم‎ 

صفحه 30:
Oxia Oratory © Ota wintey te he provess oP sewrouicwatcdly cody zieg karye databases to Pod ‏عم خی‎ سا ام من لیا میج( ۲ © rede Pa ‏پل‎ vod uppioad poses ‏له لو و‎ rich, bused oa ‏و‎ ‏ام‎ (lero, fob re, <x, «.) rod post history © Predet Po paters of phowe cakes ard usage ty Hel) ty be Prachi ۲ ‏سامت ین‎ of predvioa weckeniows: ۰ ‏یه‎ ‎١ ‏مان ما تن و مسجو‎ ches ‏و۱ ماه مایت ۰ بصن میج‎ ۰ ‏لب م6‎ ١ ven a set of cropper Por oct ‏حصا‎ Pucctos, pred the Puce ‏ای ور زر و و قح‎ سا0 لح 0 لا سواه 1 همه 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 31:
Octa Orctary (Coc) © ied books thot are oPtea bousht by “sia” oustewers. IP a eur suk custower buys var suck book, sucgeet the vers tov. مصحی تس و ‎First step‏ وه لح سا رو یه و ی لو موه وا موه وتا یی بط ‎Ohsters‏ © هد ره وه و و واه موب وی لاو ‎x).‏ * لس سس * Ortertica oP ‏اما مت وان‎ it deteviiog epidewics Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 16.00 ©Sbervehnts, Cork ced Cnakershe 

صفحه 32:
حطی() موس ۲ ‏نج موه پا حط وان‎ objerts to cheses. ۱ eurcere oppiood, should ke or she be hessiied us low risk, wedi risk or high risk? © Ops ficaiog rules Por cbove exanple could use a vere, of dota, suck a eduuived level, sokry, ۴, ‏عاج‎ 0 < سس( لمع صصص 2 و ۵ مس ۲ © ‎P.oredt = excetect‏ = ‎ood‏ اس = ‎P.deqer‏ ,۵ مس ۲ و (۲2,۵۵۵۵ ک وس ۵ امه 96,000 < عسمسس.) ‎P.credt = yood‏ >= ‎© Ques oe wt ‏و تم روج‎ wy be sowe wischssficaic © Ohkssficaiva nies con be shows cowponiy os 0 devivd ree. ‎Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏سا0 لح 0 لا سواه 1 موص‎ 

صفحه 33:
doctorate 0 excellent ©Sbervehnts, Cork ced Cnakershe bachelors _\\masters bad average good © bad O average Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏و66‎ 

صفحه 34:
+ ee ea ۲ ‏ناد مدص‎ a dota sexrple to whick the ‏سم لاه سا مشاه‎ © Greed top dows ‏ومیل اه موب‎ ees. © @uck itera ade oP the tree portiows the dota ‏ممم لحصما وحصي صم‎ ‏بو‎ virtbue, cod portizotag vondiiog Por te ode © Let xk! ۱ dd (or wont) oF he tewe ot he onde beloan to the save ches, or ‏مج روطب لو وا ما را ام‎ ‎woe ©Sbervehnts, Cork ced Cnakershe‏ 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 35:
@vsi Cpls BE Rick best ainbates oad combiows oa hick i parties Bl Dhe pany of a vet G oF traci Ket be ‏رقم لح‎ kt severd ways. © ‏ی هه اس جا - صحصمك خانم اه نس‎ = |], 1 1 2717 ‏جه لايك جا تسم خا ومح وز‎ [ ‏بت‎ )۵( - 4۶ 5 21 ۱ © Whoo dl ‏هس )سا اه و و مه مات‎ 0 © ‏ج) مجمی سم‎ ) ( ۵ ‏وج اه سوت ماه مه‎ سا0 لح 0 لا سواه 1 666 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 36:
+ @vst Opts (Ova) Bl coher weer of puny he ‏مج رو‎ whick ‏ام سا‎ cor ۹ euro )0( > - 09, 24 7 © hea a set )8 ‏صط لأدد جز‎ tiple vets (i, 120, ©, ..., ‏رع‎ we coo wear the purty oF the resutoat set oP sets os! ۲ ع 5 ‎=f burity (S)‏ )@ ری ره رم )15 = 1 BOD he iPorwaiiva yaa due ty poricukar split oP G icin G, 1 2 0, © ‏رت‎ ‏رگا رک 6) م1‎ eon G) = AG) — rr (Gy. Ge, G) Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏سا0 لح 0 لا سواه 1 موم‎ 

صفحه 37:
+ @vxt Opts (Ove) ۲ ‏مجك(‎ of “csi” oP a opi (G{G, @ ‏زرك‎ 2 Shog Ist APoreraierootest (G, {Oy Gay ‏و - - ((ز‎ —— log, 5-5 11S 151 1" ‏ات مسد بون‎ GAG, Gay... G ‏مرکا ,6) مس سم م1‎ Gay 2, GS) ۱ ‎wor‏ 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 38:
+ Pradoq Bost Opty 11 Outeepried ‏:له لح مب ان اف‎ ‏رویط(‎ sph, oe obi ‏ات و‎ chee © iar opt iy oll poseble bredkup of voker tly tur sets, ord pick the best Bl Oui evoked oirbuter (cat be sorted ia woeaciePul order) ۶ ‏توق‎ ‎» Gon chew, ‏مر لاو هس مت بو‎ ~ Ox, Picker ov (, (0, 6, 66, ‏,0ك هل‎ > 00, > © (۳ he ucke that quer best oplt © ‏تاد روش‎ ۵ ‏ارت تما ان بویت‎ x ke soxve ute hur ruchy enardeot Saal Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 666 1 ‏سا0 لح 0 لا سواه‎ 

صفحه 39:
+ OeosiowTree Oowirwioa @lyprikkw Orovedee Crow ree (G) ‏مس‎ )۵(: () مسد" جسم حدصت ‎[GO] <d,) bea‏ عورة < ( 8) روصم )ع ۳9 و ‎B;‏ ات مه ططاره ار ‎Ose best sph Porn (aorves ol airbus) iy portion‏ ‎Cri G, ce‏ © دسم ‎anton (G);‏ r Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏سا0 لح 0 لا سواه 1 و66‎ 

صفحه 40:
+ Ober Types of ‏سج ان‎ BH Wr wal wet vkeoPers ory otded i other uichueure unl ore ont evened here ۳ ‏و‎ chooPers we Owes teorew, whick swe ‏(صاع لد اهام - (ك اجام‎ ‏(9)م‎ ‎where ‏عم‎ | 1) = probblay oF ketene ‏مسال‎ ts okie ‏رك‎ ‎P (| ©) = ‏بك اه سول مه ری ط رام‎ ‏له ره سا اه موی ط رومام = ( عم‎ p (d) = probubly ‏سول‎ Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wero ©Sbervehnts, Cork ced Cnakershe 

صفحه 41:
+ Ouive Oqevtea OkeePers وس لاه ‎Daves‏ ۷ (0۱۰) همسجم ۶ 0 اه اه ۲ سوه اج ۱ و لس سا مه ( )مر © جع ات سح سح دكات كدوجو 0 سام ,عاص سا زار و ۲ وت راما اجه شا لول ‎la)". eld) 2)‏ 4( * لو اهام - لو | 0)ام ‎Cork of he p (d| 0) coo be epikosted Prow u kistoyren va dudves Por‏ © سس پر ‎۱ the Ristoqracy is ‏تم‎ Proc ‏هه روا عا‎ ۱ otirbutes are wore expeepive iy ooepuie ued store ‎Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 108.00 ©Sbervehnts, Cork ced Cnakershe 

صفحه 42:
وزووو ین ون © Rewessiva ‏و تاه مت بانط‎ vohue, roker theca ches. BE Oce uny 6 kPer coe ‏فصا جاص يد ب رک رک رنه‎ Ye qty Gta et. tag X, Bho park ot keeor poh i ood ‏ا رسي‎ © ‏ری و‎ he process of ‏يلم"‎ curve tha Pit the dota te ‏عنصت اجات مح‎ Parag. 18 ‏ج21‎ Ai voy ody be approxi © bern of ‏او‎ data, or © brome he rektiocshiy t vet excl ‏مخصيامم د‎ ۲ ‏ی مود‎ 7 Pied coe ‏ند‎ ur the best poonible Pa. Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wee ©Sbervehnts, Cork ced Cnakershe 

صفحه 43:
دج ی + ۲ Retal shops ore ‏ولا مها سوه و لو ماه‎ tiews thot people bw. © Gowevee why buys bread is quie thely dev to buy wil © © persva whe bout the book Dutabusr Gpstew Orurepts te quite tkely ober to buy the book Operatcy Gustew Orwepts. © Ossurtdives tePorwoiva cot be used to severd ways. © Gy. whoo asiower buys 0 poricder book, oc colo shop way suggest sussuciated books. © Qsevcntod res: bread = ‏لب‎ OB-Crwepts, OG-Oruepts = Detworks © bePt head side: odievedeat, right hucd side! pease quect © ‏موه و۲‎ rue ust hove oo ussecidied populaiicg; the popukaioe cveststs oF ‏واه اه بو و‎ » xx, ek rxexeton (ode) of ‏و‎ shop fs oa ketnre, ond the ort oP ol ‏سیم ات سیر‎ ‎wero ©Sbervehnts, Cork ced Cnakershe‏ 6 ,0۷06 بط 9 - مومسم 6 ی 

صفحه 44:
+ Ovevokion Rube (Ovu.) ی لس من و اس بو ,اوه لس و ما ‎Ques‏ ۱ مس سا ماه مج اج و مان هسوسو ۲ ‎nde.‏ جر مسج ‎ad the‏ © Cox. suppose voy 0.000 pervect of ol purchoses fochide wil ord soreurddvers. “Nhe support Por the nue ts wil => soreudhivers is law. جا لحم عمجت هط ‎how cPted he poosequed is‏ اه مسجت وت و0 ۲ ‎ine.‏ © Ox. ke ne bread = wilt kes ‏و سوه و‎ OO percent FOO pervedi of the purchases trot focide bread uy tockide ils. Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wee ©Sbervehnts, Cork ced Cnakershe 

صفحه 45:
+ Prachay Osevonion Ques Be oe cecerdly od) ieresied ia wsspcidiva niles wil reesvoubly high support (€.9. support oP O% or recter) ۲ ‏اسب سمش‎ ۱۳ dl posse sets oP relevent tec. ‎pucker of‏ مس ‎Brad te supper (le. ont how ony‏ موه و ‎tows tn the or).‏ ‎bere teweet! sets wih =PPreuly kids support ‏تمهت مج جا مه توا بو‎ (Crow tewset @yerde he nlc @-{b} 2b Por ewk bE @. ۷ Capper of nike = spent (@). % OnePikewe of nde = support (P ) | support (B - {}) ‎Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wero ©Sbervehnts, Cork ced Cnakershe 

صفحه 46:
Praag Gupport © Osterxoice support oF tewsrts vic o stage poss oo set oP iresurioas © bere tewsrts! sets wit o high coud of the ead oP the pos BP weer wt euch ty hold ol counts Por ofl ‏ی و‎ williple passes, ‏بط‎ ‏اس وا ها موه رل‎ poss. ۲ ‏من‎ Oure on teweet tr eric se pret ‏صمي‎ be peter, Phe prior tevkokne 7 Pied hirge tere! © Poss 0: cowl support oF all sets its just (itew. Blxoicate those tiews wits bw purport ‎oP ftews suck trot ol ts Hitec subsets ore kore‏ بو ره تال ‎Puss‏ و جج للم ‎Orart support oP ofl‏ * ‎Gtop Phere ore ow candidates:‏ * ‎ ‎ts cand (support) ‏ها‎ 1 ‎wero ©Sbervehnts, Cork ced Cnakershe‏ 6 ,0۷06 بط 9 - مومسم 6 ی ‎ 

صفحه 47:
6 Types oP @se00ntiow § @cstc ossertaicg ndes hove severd bevtaiccs ۲ Oevisiogs Pow the experted probubiiiy are wore ‏ها‎ © C.y. Pwo people porches breed, oad woop people ‏اس‎ veredl, ‏د عشي‎ Pew wold be expevied to ‏یم‎ bots © Oe we terested ta postive us welles oetnive corrections between sets oP tems ۱ @ostive corretticd: co-pocun cope ts higher thoc predicted ۲ ‏موی ماه مهم(‎ ts lower trac predicted © Gequews ‏اه | یه‎ ۶ ‏حلص مه رم من عمط بیج‎ tO days © ‏تون‎ Pro tewpord puteros © G.y, devicion Brow « steady growls 8 ) sdles oP wtoter wear yo dowd to Sucre * Oot surprsiy, pot oF a koowa poterc. * book ‏قرط و‎ Brow uddue predicted usta post putlerc Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wer ©Sbervehnts, Cork ced Cnakershe 

صفحه 48:
Ohperiay «8 ‏رامش نصا‎ Prrdoy clusters of points fo the yived data suck thot stakes price ‏عخصام ی با ما با‎ روت یحو وا وی سل بو لاو چا نون ‎B®‏ ۶ Grow ports tie roots (Por a ued f) suck thot he average dooce ‏عمج‎ ‎Beano the ‏امس‎ of ‏اه مت له‎ و اس و وی و مه ماه روا بای سس لسمین) ۲ © ‏موه وه ی لو‎ dstoue betwerd every por oP potas ino hoster Bice bee sited extrosively ‏مدل امد من انحا بوصاصاماد جا‎ ete © ‏مد موه بت مت‎ of cistertay tevhaques that roa hoa very kere dott vee © x. he @irck chetertery chorion (coor okerty) Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wwe ©Sbervehnts, Cork ced Cnakershe 

صفحه 49:
Werachicd Ohsteriag Breet Brey barca cheer © (ke word ‏سوه مها بجوو كاسما‎ oot wean 9 predoioa ‏صم ممم‎ ۵ ست سوت مس تما ‎systews (e.q. ‘Vchow, wore va this kter)‏ رووتسط موه تطامووت م0 ۲ ‎Oggiveraive chsiertay deprihes‏ لا ‎und oy oo‏ ماه جوا ما مضه وت اه ما واه لو لش © اد اه را ۱ ‎rePiae (breuk) cheters iio sadly‏ ۱ Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO wero ©Sbervehnts, Cork ced Cnakershe 

صفحه 50:
+ 0 1 ‏سا مها ماب رصان‎ destqued to hod very harge dotasets © Cy, the @irch ‏اه‎ ‎© Das ites! use oo evewory rer te store poids thot are betoy chistered © Sosert ports vee oto fre ‏جه کاس ار سس هر )سا ما‎ ‏صاه مش‎ P is bees tho soxwe 3 deter away 0B ere are wore lee odes thon Pil io wewory, werye existing chsters thot are dose to rock other ۴ )۵ ۳ cud of First pass we yet ohare oueber of chisters of the leuwes oP the R-tree * Derg chisters to reduce te anober of chester Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO ‏سا0 لح 0 لا سواه 1 همه‎ 

صفحه 51:
$ave ern © Gok predict ut movies lboobesl ... per ray be tote dict, oro oe ممصم و تا موی وا وحمب ام سم ام ‎Ober people wi‏ © © Dhe prePerewwes oP suck people Por a caw ‏.لالج‎ ‎© Ove ‏اه لو من لیا اس‎ © Chester people va the busts oP prePerewer Por woview © Dhea chester wovies vu the boss oP bet thed by the sae chisters oP peuple © @ysic chister peuple bused vu their prePereuves Por (he ‏ات را‎ cheters of) coves © Repeat cbove il equlbrice ۲ Obie ‏ماس‎ fs vo etre oP poloboraive Pilertay, where vers ‏مان‎ ‎fo the task oP Pthertoy tePoroaica to Ped iPorcaica ‏او چاو‎ wou Oxsdrer Gyetre Oncewpte - 0" Briar, xq GO, OOOO 

صفحه 52:
+ Oter Tyger oP Oracg © Text wictey: upphicaica oP dota cinta to textud douse © chester Deb pages ty Prod retiied pages © cheter payer 3 wer has visited to organize their vist history © ches) Orb paces cuirraicdy cio o Deb devin © Oct vevalzaton systews help users exavice hrye uokeves of dott ood detect ‏بلاسجاب صحصادم‎ مه و وم كلما ان جتسججمه جيمها جلحصب براسجاب 055 © ام و ‎a‏ لو ‎very‏ وه ما ۶ سا0 لح 0 لا سواه 1 666 6 ,0۷06 بط 9 - مومسم 6 ی 

Chapter 18: Data Analysis and Mining Database System Concepts ©Silberschatz, Korth and Sudarshan See www.db-book.com for conditions on re-use Chapter 18: Data Analysis and Mining  Decision Support Systems  Data Analysis and OLAP  Data Warehousing  Data Mining Database System Concepts - 5th Edition, Aug 26, 2005 18.2 ©Silberschatz, Korth and Sudarshan Decision Support Systems  Decision-support systems are used to make business decisions, often based on data collected by on-line transaction-processing systems.  Examples of business decisions:   What items to stock?  What insurance premium to change?  To whom to send advertisements? Examples of data used for making decisions  Retail sales transaction details  Customer profiles (income, age, gender, etc.) Database System Concepts - 5th Edition, Aug 26, 2005 18.3 ©Silberschatz, Korth and Sudarshan Decision-Support Systems: Overview  Data analysis tasks are simplified by specialized tools and SQL extensions   Example tasks  For each product category and each region, what were the total sales in the last quarter and how do they compare with the same quarter last year  As above, for each product category and each customer category Statistical analysis packages (e.g., : S++) can be interfaced with databases  Statistical analysis is a large field, but not covered here  Data mining seeks to discover knowledge automatically in the form of statistical rules and patterns from large databases.  A data warehouse archives information gathered from multiple sources, and stores it under a unified schema, at a single site.  Important for large businesses that generate data from multiple divisions, possibly at multiple sites  Data may also be purchased externally Database System Concepts - 5th Edition, Aug 26, 2005 18.4 ©Silberschatz, Korth and Sudarshan Data Analysis and OLAP  Online Analytical Processing (OLAP)   Interactive analysis of data, allowing data to be summarized and viewed in different ways in an online fashion (with negligible delay) Data that can be modeled as dimension attributes and measure attributes are called multidimensional data.   Measure attributes  measure some value  can be aggregated upon  e.g. the attribute number of the sales relation Dimension attributes  define the dimensions on which measure attributes (or aggregates thereof) are viewed  e.g. the attributes item_name, color, and size of the sales relation Database System Concepts - 5th Edition, Aug 26, 2005 18.5 ©Silberschatz, Korth and Sudarshan Cross Tabulation of sales by item-name and color  The table above is an example of a cross-tabulation (cross-tab), also referred to as a pivot-table.  Values for one of the dimension attributes form the row headers  Values for another dimension attribute form the column headers  Other dimension attributes are listed on top  Values in individual cells are (aggregates of) the values of the dimension attributes that specify the cell. Database System Concepts - 5th Edition, Aug 26, 2005 18.6 ©Silberschatz, Korth and Sudarshan Relational Representation of Cross-tabs  Cross-tabs can be represented as relations  We use the value all is used to represent aggregates  The SQL:1999 standard actually uses null values in place of all despite confusion with regular null values Database System Concepts - 5th Edition, Aug 26, 2005 18.7 ©Silberschatz, Korth and Sudarshan Data Cube  A data cube is a multidimensional generalization of a cross-tab  Can have n dimensions; we show 3 below  Cross-tabs can be used as views on a data cube Database System Concepts - 5th Edition, Aug 26, 2005 18.8 ©Silberschatz, Korth and Sudarshan Online Analytical Processing  Pivoting: changing the dimensions used in a cross-tab is called  Slicing: creating a cross-tab for fixed values only  Sometimes called dicing, particularly when values for multiple dimensions are fixed.  Rollup: moving from finer-granularity data to a coarser granularity  Drill down: The opposite operation - that of moving from coarser-granularity data to finer-granularity data Database System Concepts - 5th Edition, Aug 26, 2005 18.9 ©Silberschatz, Korth and Sudarshan Hierarchies on Dimensions  Hierarchy on dimension attributes: lets dimensions to be viewed at different levels of detail  E.g. the dimension DateTime can be used to aggregate by hour of day, date, day of week, month, quarter or year Database System Concepts - 5th Edition, Aug 26, 2005 18.10 ©Silberschatz, Korth and Sudarshan Cross Tabulation With Hierarchy  Cross-tabs can be easily extended to deal with hierarchies  Can drill down or roll up on a hierarchy Database System Concepts - 5th Edition, Aug 26, 2005 18.11 ©Silberschatz, Korth and Sudarshan OLAP Implementation  The earliest OLAP systems used multidimensional arrays in memory to store data cubes, and are referred to as multidimensional OLAP (MOLAP) systems.  OLAP implementations using only relational database features are called relational OLAP (ROLAP) systems  Hybrid systems, which store some summaries in memory and store the base data and other summaries in a relational database, are called hybrid OLAP (HOLAP) systems. Database System Concepts - 5th Edition, Aug 26, 2005 18.12 ©Silberschatz, Korth and Sudarshan OLAP Implementation (Cont.)  Early OLAP systems precomputed all possible aggregates in order to provide online response  Space and time requirements for doing so can be very high   2n combinations of group by It suffices to precompute some aggregates, and compute others on demand from one of the precomputed aggregates  Can compute aggregate on (item-name, color) from an aggregate on (itemname, color, size) – For all but a few “non-decomposable” aggregates such as median – is cheaper than computing it from scratch  Several optimizations available for computing multiple aggregates  Can compute aggregate on (item-name, color) from an aggregate on (item-name, color, size)  Can compute aggregates on (item-name, color, size), (item-name, color) and (item-name) using a single sorting of the base data Database System Concepts - 5th Edition, Aug 26, 2005 18.13 ©Silberschatz, Korth and Sudarshan Extended Aggregation in SQL:1999  The cube operation computes union of group by’s on every subset of the specified attributes  E.g. consider the query select item-name, color, size, sum(number) from sales group by cube(item-name, color, size) This computes the union of eight different groupings of the sales relation: { (item-name, color, size), (item-name, color), (item-name, size), (color, size), (item-name), (color), (size), ()} where ( ) denotes an empty group by list.  For each grouping, the result contains the null value for attributes not present in the grouping. Database System Concepts - 5th Edition, Aug 26, 2005 18.14 ©Silberschatz, Korth and Sudarshan Extended Aggregation (Cont.)  Relational representation of cross-tab that we saw earlier, but with null in place of all, can be computed by select item-name, color, sum(number) from sales group by cube(item-name, color)  The function grouping() can be applied on an attribute  Returns 1 if the value is a null value representing all, and returns 0 in all other cases. select item-name, color, size, sum(number), grouping(item-name) as item-name-flag, grouping(color) as color-flag, grouping(size) as size-flag, from sales group by cube(item-name, color, size)  Can use the function decode() in the select clause to replace such nulls by a value such as all  E.g. replace item-name in first query by decode( grouping(item-name), 1, ‘all’, item-name) Database System Concepts - 5th Edition, Aug 26, 2005 18.15 ©Silberschatz, Korth and Sudarshan Extended Aggregation (Cont.)  The rollup construct generates union on every prefix of specified list of attributes  E.g. select item-name, color, size, sum(number) from sales group by rollup(item-name, color, size) Generates union of four groupings: { (item-name, color, size), (item-name, color), (item-name), ( ) }  Rollup can be used to generate aggregates at multiple levels of a hierarchy.  E.g., suppose table itemcategory(item-name, category) gives the category of each item. Then select category, item-name, sum(number) from sales, itemcategory where sales.item-name = itemcategory.item-name group by rollup(category, item-name) would give a hierarchical summary by item-name and by category. Database System Concepts - 5th Edition, Aug 26, 2005 18.16 ©Silberschatz, Korth and Sudarshan Extended Aggregation (Cont.)  Multiple rollups and cubes can be used in a single group by clause   Each generates set of group by lists, cross product of sets gives overall set of group by lists E.g., select item-name, color, size, sum(number) from sales group by rollup(item-name), rollup(color, size) generates the groupings {item-name, ()} X {(color, size), (color), ()} = { (item-name, color, size), (item-name, color), (item-name), (color, size), (color), ( ) } Database System Concepts - 5th Edition, Aug 26, 2005 18.17 ©Silberschatz, Korth and Sudarshan Ranking  Ranking is done in conjunction with an order by specification.  Given a relation student-marks(student-id, marks) find the rank of each student. select student-id, rank( ) over (order by marks desc) as s-rank from student-marks  An extra order by clause is needed to get them in sorted order select student-id, rank ( ) over (order by marks desc) as s-rank from student-marks order by s-rank  Ranking may leave gaps: e.g. if 2 students have the same top mark, both have rank 1, and the next rank is 3  dense_rank does not leave gaps, so next dense rank would be 2 Database System Concepts - 5th Edition, Aug 26, 2005 18.18 ©Silberschatz, Korth and Sudarshan Ranking (Cont.)  Ranking can be done within partition of the data.  “Find the rank of students within each section.” select student-id, section, rank ( ) over (partition by section order by marks desc) as sec-rank from student-marks, student-section where student-marks.student-id = student-section.student-id order by section, sec-rank  Multiple rank clauses can occur in a single select clause  Ranking is done after applying group by clause/aggregation Database System Concepts - 5th Edition, Aug 26, 2005 18.19 ©Silberschatz, Korth and Sudarshan Ranking (Cont.)  Other ranking functions:  percent_rank (within partition, if partitioning is done)  cume_dist (cumulative distribution)    fraction of tuples with preceding values row_number (non-deterministic in presence of duplicates) SQL:1999 permits the user to specify nulls first or nulls last select student-id, rank ( ) over (order by marks desc nulls last) as s-rank from student-marks Database System Concepts - 5th Edition, Aug 26, 2005 18.20 ©Silberschatz, Korth and Sudarshan Ranking (Cont.)  For a given constant n, the ranking the function ntile(n) takes the tuples in each partition in the specified order, and divides them into n buckets with equal numbers of tuples.  E.g.: select threetile, sum(salary) from ( select salary, ntile(3) over (order by salary) as threetile from employee) as s group by threetile Database System Concepts - 5th Edition, Aug 26, 2005 18.21 ©Silberschatz, Korth and Sudarshan Windowing  Used to smooth out random variations.  E.g.: moving average: “Given sales values for each date, calculate for each date the average of the sales on that day, the previous day, and the next day”  Window specification in SQL:  Given relation sales(date, value) select date, sum(value) over (order by date between rows 1 preceding and 1 following) from sales  Examples of other window specifications:  between rows unbounded preceding and current  rows unbounded preceding  range between 10 preceding and current row   All rows with values between current row value –10 to current value range interval 10 day preceding  Not including current row Database System Concepts - 5th Edition, Aug 26, 2005 18.22 ©Silberschatz, Korth and Sudarshan Windowing (Cont.)  Can do windowing within partitions  E.g. Given a relation transaction (account-number, date-time, value), where value is positive for a deposit and negative for a withdrawal  “Find total balance of each account after each transaction on the account” select account-number, date-time, sum (value ) over (partition by account-number order by date-time rows unbounded preceding) as balance from transaction order by account-number, date-time Database System Concepts - 5th Edition, Aug 26, 2005 18.23 ©Silberschatz, Korth and Sudarshan Data Warehousing  Data sources often store only current data, not historical data  Corporate decision making requires a unified view of all organizational data, including historical data  A data warehouse is a repository (archive) of information gathered from multiple sources, stored under a unified schema, at a single site  Greatly simplifies querying, permits study of historical trends  Shifts decision support query load away from transaction processing systems Database System Concepts - 5th Edition, Aug 26, 2005 18.24 ©Silberschatz, Korth and Sudarshan Data Warehousing Database System Concepts - 5th Edition, Aug 26, 2005 18.25 ©Silberschatz, Korth and Sudarshan Design Issues   When and how to gather data  Source driven architecture: data sources transmit new information to warehouse, either continuously or periodically (e.g. at night)  Destination driven architecture: warehouse periodically requests new information from data sources  Keeping warehouse exactly synchronized with data sources (e.g. using twophase commit) is too expensive  Usually OK to have slightly out-of-date data at warehouse  Data/updates are periodically downloaded form online transaction processing (OLTP) systems. What schema to use  Schema integration Database System Concepts - 5th Edition, Aug 26, 2005 18.26 ©Silberschatz, Korth and Sudarshan More Warehouse Design Issues   Data cleansing  E.g. correct mistakes in addresses (misspellings, zip code errors)  Merge address lists from different sources and purge duplicates How to propagate updates   Warehouse schema may be a (materialized) view of schema from data sources What data to summarize  Raw data may be too large to store on-line  Aggregate values (totals/subtotals) often suffice  Queries on raw data can often be transformed by query optimizer to use aggregate values Database System Concepts - 5th Edition, Aug 26, 2005 18.27 ©Silberschatz, Korth and Sudarshan Warehouse Schemas  Dimension values are usually encoded using small integers and mapped to full values via dimension tables  Resultant schema is called a star schema  More complicated schema structures  Snowflake schema: multiple levels of dimension tables  Constellation: multiple fact tables Database System Concepts - 5th Edition, Aug 26, 2005 18.28 ©Silberschatz, Korth and Sudarshan Data Warehouse Schema Database System Concepts - 5th Edition, Aug 26, 2005 18.29 ©Silberschatz, Korth and Sudarshan Data Mining  Data mining is the process of semi-automatically analyzing large databases to find useful patterns  Prediction based on past history   Predict if a credit card applicant poses a good credit risk, based on some attributes (income, job type, age, ..) and past history  Predict if a pattern of phone calling card usage is likely to be fraudulent Some examples of prediction mechanisms:  Classification   Given a new item whose class is unknown, predict to which class it belongs Regression formulae  Given a set of mappings for an unknown function, predict the function result for a new parameter value Database System Concepts - 5th Edition, Aug 26, 2005 18.30 ©Silberschatz, Korth and Sudarshan Data Mining (Cont.)  Descriptive Patterns  Associations   Associations may be used as a first step in detecting causation   Find books that are often bought by “similar” customers. If a new such customer buys one such book, suggest the others too. E.g. association between exposure to chemical X and cancer, Clusters  E.g. typhoid cases were clustered in an area surrounding a contaminated well  Detection of clusters remains important in detecting epidemics Database System Concepts - 5th Edition, Aug 26, 2005 18.31 ©Silberschatz, Korth and Sudarshan Classification Rules  Classification rules help assign new objects to classes.   E.g., given a new automobile insurance applicant, should he or she be classified as low risk, medium risk or high risk? Classification rules for above example could use a variety of data, such as educational level, salary, age, etc.   person P, P.degree = masters and P.income > 75,000  P.credit = excellent   person P, P.degree = bachelors and (P.income  25,000 and P.income  75,000)  P.credit = good  Rules are not necessarily exact: there may be some misclassifications  Classification rules can be shown compactly as a decision tree. Database System Concepts - 5th Edition, Aug 26, 2005 18.32 ©Silberschatz, Korth and Sudarshan Decision Tree Database System Concepts - 5th Edition, Aug 26, 2005 18.33 ©Silberschatz, Korth and Sudarshan Construction of Decision Trees  Training set: a data sample in which the classification is already known.  Greedy top down generation of decision trees.  Each internal node of the tree partitions the data into groups based on a partitioning attribute, and a partitioning condition for the node  Leaf node:  all (or most) of the items at the node belong to the same class, or  all attributes have been considered, and no further partitioning is possible. Database System Concepts - 5th Edition, Aug 26, 2005 18.34 ©Silberschatz, Korth and Sudarshan Best Splits  Pick best attributes and conditions on which to partition  The purity of a set S of training instances can be measured quantitatively in several ways.   Notation: number of classes = k, number of instances = |S|, fraction of instances in class i = pi. The Gini measure of purity is defined as [ Gini (S) = 1k-  p2i i- 1  When all instances are in a single class, the Gini value is 0  It reaches its maximum (of 1 –1 /k) if each class the same number of instances. Database System Concepts - 5th Edition, Aug 26, 2005 18.35 ©Silberschatz, Korth and Sudarshan Best Splits (Cont.)  Another measure of purity is the entropy measure, which is defined as k entropy (S) = – pilog2 i- 1 pi  When a set S is split into multiple sets Si, I=1, 2, …, r, we can measure the purity of the resultant set of sets as: r |Si| purity(S1, S2, ….., Sr) =  purity (Si) i= |S| 1  The information gain due to particular split of S into S i, i = 1, 2, …., r Information-gain (S, {S1, S2, …., Sr) = purity(S ) – purity (S1, S2, … Sr) Database System Concepts - 5th Edition, Aug 26, 2005 18.36 ©Silberschatz, Korth and Sudarshan Best Splits (Cont.)  Measure of “cost” of a split: Information-content (S, {S1, S2, ….., Sr})) = –  r |S | i i- 1 |S|  log2 |Si| |S| Information-gain ratio = Information-gain (S, {S1, S2, ……, Sr}) Information-content (S, {S1, S2, ….., Sr})  The best split is the one that gives the maximum information gain ratio Database System Concepts - 5th Edition, Aug 26, 2005 18.37 ©Silberschatz, Korth and Sudarshan Finding Best Splits   Categorical attributes (with no meaningful order):  Multi-way split, one child for each value  Binary split: try all possible breakup of values into two sets, and pick the best Continuous-valued attributes (can be sorted in a meaningful order)   Binary split:  Sort values, try each as a split point – E.g. if values are 1, 10, 15, 25, split at 1,  10,  15  Pick the value that gives best split Multi-way split:  A series of binary splits on the same attribute has roughly equivalent effect Database System Concepts - 5th Edition, Aug 26, 2005 18.38 ©Silberschatz, Korth and Sudarshan Decision-Tree Construction Algorithm Procedure GrowTree (S ) Partition (S ); Procedure Partition (S) if ( purity (S ) > p or |S| < s ) then return; for each attribute A evaluate splits on attribute A; Use best split found (across all attributes) to partition S into S1, S2, …., Sr, for i = 1, 2, ….., r Partition (Si ); Database System Concepts - 5th Edition, Aug 26, 2005 18.39 ©Silberschatz, Korth and Sudarshan Other Types of Classifiers  Neural net classifiers are studied in artificial intelligence and are not covered here  Bayesian classifiers use Bayes theorem, which says where p (c j | d ) = p (d | c j ) p (c j ) p(d) p (cj | d ) = probability of instance d being in class cj, p (d | cj ) = probability of generating instance d given class cj, p (cj ) = probability of occurrence of class cj, and p (d ) = probability of instance d occuring Database System Concepts - 5th Edition, Aug 26, 2005 18.40 ©Silberschatz, Korth and Sudarshan Naïve Bayesian Classifiers   Bayesian classifiers require  computation of p (d | cj )  precomputation of p (cj )  p (d ) can be ignored since it is the same for all classes To simplify the task, naïve Bayesian classifiers assume attributes have independent distributions, and thereby estimate p (d | cj) = p (d1 | cj ) * p (d2 | cj ) * ….* (p (dn | cj )  Each of the p (di | cj ) can be estimated from a histogram on di values for each class cj   the histogram is computed from the training instances Histograms on multiple attributes are more expensive to compute and store Database System Concepts - 5th Edition, Aug 26, 2005 18.41 ©Silberschatz, Korth and Sudarshan Regression  Regression deals with the prediction of a value, rather than a class.  Given values for a set of variables, X1, X2, …, Xn, we wish to predict the value of a variable Y.  One way is to infer coefficients a0, a1, a1, …, an such that Y = a0 + a1 * X1 + a2 * X2 + … + an * Xn  Finding such a linear polynomial is called linear regression.    In general, the process of finding a curve that fits the data is also called curve fitting. The fit may only be approximate  because of noise in the data, or  because the relationship is not exactly a polynomial Regression aims to find coefficients that give the best possible fit. Database System Concepts - 5th Edition, Aug 26, 2005 18.42 ©Silberschatz, Korth and Sudarshan Association Rules   Retail shops are often interested in associations between different items that people buy.  Someone who buys bread is quite likely also to buy milk  A person who bought the book Database System Concepts is quite likely also to buy the book Operating System Concepts. Associations information can be used in several ways.   E.g. when a customer buys a particular book, an online shop may suggest associated books. Association rules: bread  milk DB-Concepts, OS-Concepts  Networks  Left hand side: antecedent, right hand side: consequent  An association rule must have an associated population; the population consists of a set of instances  E.g. each transaction (sale) at a shop is an instance, and the set of all transactions is the population Database System Concepts - 5th Edition, Aug 26, 2005 18.43 ©Silberschatz, Korth and Sudarshan Association Rules (Cont.)  Rules have an associated support, as well as an associated confidence.  Support is a measure of what fraction of the population satisfies both the antecedent and the consequent of the rule.   E.g. suppose only 0.001 percent of all purchases include milk and screwdrivers. The support for the rule is milk  screwdrivers is low. Confidence is a measure of how often the consequent is true when the antecedent is true.  E.g. the rule bread  milk has a confidence of 80 percent if 80 percent of the purchases that include bread also include milk. Database System Concepts - 5th Edition, Aug 26, 2005 18.44 ©Silberschatz, Korth and Sudarshan Finding Association Rules  We are generally only interested in association rules with reasonably high support (e.g. support of 2% or greater)  Naïve algorithm 1. Consider all possible sets of relevant items. 2. For each set find its support (i.e. count how many transactions purchase all items in the set).  3. Large itemsets: sets with sufficiently high support Use large itemsets to generate association rules. 1. From itemset A generate the rule A - {b } b for each b  A.  Support of rule = support (A).  Confidence of rule = support (A ) / support (A - {b }) Database System Concepts - 5th Edition, Aug 26, 2005 18.45 ©Silberschatz, Korth and Sudarshan Finding Support  Determine support of itemsets via a single pass on set of transactions  Large itemsets: sets with a high count at the end of the pass  If memory not enough to hold all counts for all itemsets use multiple passes, considering only some itemsets in each pass.  Optimization: Once an itemset is eliminated because its count (support) is too small none of its supersets needs to be considered.  The a priori technique to find large itemsets:  Pass 1: count support of all sets with just 1 item. Eliminate those items with low support  Pass i: candidates: every set of i items such that all its i-1 item subsets are large  Count support of all candidates  Stop if there are no candidates Database System Concepts - 5th Edition, Aug 26, 2005 18.46 ©Silberschatz, Korth and Sudarshan Other Types of Associations  Basic association rules have several limitations  Deviations from the expected probability are more interesting   E.g. if many people purchase bread, and many people purchase cereal, quite a few would be expected to purchase both  We are interested in positive as well as negative correlations between sets of items Positive correlation: co-occurrence is higher than predicted  Negative correlation: co-occurrence is lower than predicted Sequence associations / correlations    E.g. whenever bonds go up, stock prices go down in 2 days Deviations from temporal patterns  E.g. deviation from a steady growth  E.g. sales of winter wear go down in summer  Not surprising, part of a known pattern.  Look for deviation from value predicted using past patterns Database System Concepts - 5th Edition, Aug 26, 2005 18.47 ©Silberschatz, Korth and Sudarshan Clustering  Clustering: Intuitively, finding clusters of points in the given data such that similar points lie in the same cluster  Can be formalized using distance metrics in several ways  Group points into k sets (for a given k) such that the average distance of points from the centroid of their assigned group is minimized    Centroid: point defined by taking average of coordinates in each dimension. Another metric: minimize average distance between every pair of points in a cluster Has been studied extensively in statistics, but on small data sets  Data mining systems aim at clustering techniques that can handle very large data sets  E.g. the Birch clustering algorithm (more shortly) Database System Concepts - 5th Edition, Aug 26, 2005 18.48 ©Silberschatz, Korth and Sudarshan Hierarchical Clustering  Example from biological classification  (the word classification here does not mean a prediction mechanism) chordata mammalia leopards humans reptilia snakes crocodiles  Other examples: Internet directory systems (e.g. Yahoo, more on this later)  Agglomerative clustering algorithms   Build small clusters, then cluster small clusters into bigger clusters, and so on Divisive clustering algorithms  Start with all items in a single cluster, repeatedly refine (break) clusters into smaller ones Database System Concepts - 5th Edition, Aug 26, 2005 18.49 ©Silberschatz, Korth and Sudarshan Clustering Algorithms  Clustering algorithms have been designed to handle very large datasets  E.g. the Birch algorithm  Main idea: use an in-memory R-tree to store points that are being clustered  Insert points one at a time into the R-tree, merging a new point with an existing cluster if is less than some  distance away  If there are more leaf nodes than fit in memory, merge existing clusters that are close to each other  At the end of first pass we get a large number of clusters at the leaves of the R-tree  Merge clusters to reduce the number of clusters Database System Concepts - 5th Edition, Aug 26, 2005 18.50 ©Silberschatz, Korth and Sudarshan Collaborative Filtering    Goal: predict what movies/books/… a person may be interested in, on the basis of  Past preferences of the person  Other people with similar past preferences  The preferences of such people for a new movie/book/… One approach based on repeated clustering  Cluster people on the basis of preferences for movies  Then cluster movies on the basis of being liked by the same clusters of people  Again cluster people based on their preferences for (the newly created clusters of) movies  Repeat above till equilibrium Above problem is an instance of collaborative filtering, where users collaborate in the task of filtering information to find information of interest Database System Concepts - 5th Edition, Aug 26, 2005 18.51 ©Silberschatz, Korth and Sudarshan Other Types of Mining   Text mining: application of data mining to textual documents  cluster Web pages to find related pages  cluster pages a user has visited to organize their visit history  classify Web pages automatically into a Web directory Data visualization systems help users examine large volumes of data and detect patterns visually  Can visually encode large amounts of information on a single screen  Humans are very good a detecting visual patterns Database System Concepts - 5th Edition, Aug 26, 2005 18.52 ©Silberschatz, Korth and Sudarshan