Optimizing Your Analytic s L ife Cyc le with SAS & T e r adata Paul Se gal - T e r adata June 2017
• Ag e nda • T he Ana lytic L ife Cyc le • Co mmo n Pro b le ms • SAS & T e ra da ta so lutio ns 2
Ana lytic a l L ife Cyc le Exploration Preparation Development Deployment Deliver Explore All Prepare Build Analytic Results to Your Data Data for Models Business Analytics 3 3
Ana lytic a l L ife Cyc le • c o mb ining da ta fro m nume ro us so urc e s • ha ndling inc o nsiste nt o r no n-sta nda rdize d da ta • c le a ning dirty da ta Stage 2 • inte g ra ting da ta tha t wa s ma nua lly e nte re d • de a ling with se mi-struc ture d a nd struc ture d da ta PRE PARAT ION • c usto me r re te ntio n DE • c usto me r a ttritio n/ c hurn Stage 1 VE • ma rke ting re spo nse Stage 3 L OPME • c o nsume r lo ya lty a nd o ffe rs • wha t the d a ta lo o ks like • • wha t va ria b le s a re in the d a ta se t fra ud d e te c tio n IO T E XT ORAT • • c re d it sc o ring whe the r the re a re a ny missing NT • risk ma na g e me nt o b se rva tio ns N • ho w a re the d a ta re la te d XPL • wha t a re so me o f the d a ta pa tte rns E DE PL OYME NT • the pro b a b ility o f re spo nd ing to a pa rtic ula r pro mo tio na l o ffe r Stage 4 • the risk o f a n a pplic a nt d e fa ulting o n a lo a n • the pro pe nsity to pa y o ff a d e b t • the like liho o d a c usto me r le a ve / c hurn • the pro b a b ility to b uy a pro d uc t 4
Co mmo n Pro b le ms 5 5
T ypic a l Custo me r Cha lle ng e s “ My a na lytic a l pro c e ss runs to o slo wly * ” • • “ We spe nd to o muc h time mo ving da ta a ro und b e twe e n syste ms ” • “ I wa nt to ma ke mo re / b e tte r use o f my E DW/ da ta pla tfo rm” • “ I t ta ke s fo re ve r to e xtra c t a nd sc o re the da ta ” • “ T he re is to o muc h da ta fo r us to a na lyse ” • “ We c a n’ t b uy ne w ha rdwa re ” • “T he q ua lity o f o ur a na lytic a l mo de ls is lo we r b e c a use we sa mple – I wo rry we a re missing va lua b le se g me nts” * sc o ring / a na lysis / da ta q ua lity / da ta tra nsfo rmatio n 6
T he Ana lytic Da ta Wa re ho use LINE O F BUSINESS PRO G RAM MANAG ER ADW FINANC E SUPPLY C HAIN EDW I.T . HR 7
Da ta Ma na g e me nt: T he c rux o f the issue fa c ing yo ur a na lysts BUSI NE SS BUSI NE SS PROBL E M DE CI SI ON 80% 20% Pre pa ring to So lving so lve the pro b le m the pro b le m 8
Da ta Ma na g e me nt: SAS & T e ra da ta wo rking to c ha ng e the E q ua tio n BUSI NE SS BUSI NE SS PROBL E M DE CI SI ON 80% 20% Pre pa ring So lving the to so lve pro b le m the pro b le m 9
Ba rrie rs to the Ado ptio n o f Ana lytic s Sc a r c ity of a na lytic a l skills T he ne e d to g ro w analytic al tale nt fro m within T ools that ar e n’t r ight for the job L e arning c urve to c re ate , share and c o llab o rate Disjointe d, ine ffic ie nt wor kflow Ho w c an yo u fail fast & le arn to re fine q uic kly 10
SAS & T e r adata Solutions 1 11 1
Ana lytic a l L ife Cyc le Development Deployment Preparation Exploration ORDER ORDER NUMBER ORDER STATUS D A ORDER I TEM BACKORDERED T QUANTI TY CUSTOMER a CUSTOMER NUMBER E CUSTOMER NAME ORDER I TEM SHI PPED CUSTOMER CI TY QUANTI TY CUSTOMER POST SHI P DATE CUSTOMER ST CUSTOMER ADDR CUSTOMER PHONE I TEM CUSTOMER FAX QUANTI TY DESCRI PTI ON • Analytics • SAS Scoring • ACCESS to • Visual Accelerator Accelerator Teradata Analytics & for Teradata • SAS Model • Code Visual • SAS High- Accelerator Manager Statistics Performance • Teradata • Data Quality • Teradata Analytics Appliance for Accelerator Appliance Products SAS • Data Set for SAS • TD Appliance Builder for for SAS SAS 12
SAS Ana lytic s in T e ra da ta Da ta Mining / Hig h Info rma tio n Pre dic tive F ra ud & Risk Re po rting & Ano ma ly T e xt Ana lytic s Pe rfo rma nc e Ma na g e me nt Ana lytic s De te c tio n Visua liza tio n De te c tio n Ana lytic s I nte g ra te d E nd-to -E nd F o unda tio n DEPLOYMENT FLEXIBILITY: DESKTOP - SERVER - IN-DATABASE - IN-MEMORY - CLOUD ARCHITECTURE FLEXIBILITY: SMP - MPP - HADOOP - GRID - ESP 13
I n-Da ta b a se F unc tio na lity SAS Code Ac c e le r ator for T e r adata SAS Sc or ing Ac c e le r ator for T e r adata SAS/ Ac c e ss to T e ra da ta : • PROC DS2 • E M/ ST AT * Mo de ls • PROC APPE ND • PROC CONT E NT S Statistic al Analysis • PROC COPY SAS Ana lytic s Ac c e le ra tor for T e ra da ta • PROC DAT ASE T S Pr oc e dur e s : • PROC DE L E T E • PROC F ORMAT • PROC SCORE wo rks with • PROC CANCORR • PROC F RE Q c o e ffic ie nts fro m: • PROC CORR • PROC ME ANS • PROC F ACT OR • PROC PRINT • PROC ACE CL US • PROC PRINCOMP • PROC RANK • PROC CAL IS • PROC RE G • PROC RE PORT • PROC CANDISC • PROC SCORE • PROC SORT • PROC DISCRI M • PROC T IME SE RI E S • PROC SQL • PROC F ACT OR • PROC VARCL US • PROC SUMMARY • PROC PRINCOMP • PROC T ABUL AT E • PROC T CAL IS SAS Enterprise Miner • PROC VARCL US • PROC DMDB • PROC ORT HORE G DQ Ac c e le ra tor for • PROC DMINE • PROC QUANT RE G T e ra da ta • PROC DMRE G (L o g istic Re g re ssio n) •Ma tc h c o de • PROC RE G • Also no de s fo r Input, Sa mple , Pa rtitio n, F ilte r, •Pa rsing / Ca sing • PROC ROBUST RE G Me rg e , E xpa nd •Ge nde r/ Pa tte rn/ Ide ntific atio n a na lysis •Sta nda rdiza tio n 14
I n-Da ta b a se E xa mple : Da ta Qua lity • SAS Da ta Qua lity func tio ns po rte d to o pe ra te in-datab ase • I nvo ke d a s SQL Sto re d Pro c e dure s • Pro c e ssing le ve ra g e s pa ra lle lism o f unde rlying RDBMS • Sig nific a nt pe rfo rma nc e / thro ug hput b e ne fits • Suppo rte d func tio ns: • Ma tc hc o de g e ne ra tio n • Pa rsing • Sta nda rdiza tio n • Ca sing • Pa tte rn a na lysis • I de ntific a tio n a na lysis • Ge nde r a na lysis 15
E xplo ra tio n a nd mo de ling – wha t do e s it b ring ? Influe nc e Re la tionships Contributions E limina ting g ue sswork Inc r e ase d Da ta drive n e xplora tion with pre dic tive unde r standing a na lytic s 16
SAS Visua l Ana lytic s & Visua l Sta tistic s SAS Visua l Ana lytic s • Da ta e xplo ra tio n a nd d isc o ve ry • Distrib utio n a nd summa ry sta tistic s • Po st-mo d e l a na lysis a nd re po rting SAS Visua l Sta tistic s • Build pre d ic tio n a nd c la ssific a tio n mo d e ls • Re fine c a nd id a te mo d e ls • Co mpa re mo d e ls a nd g e ne ra te sc o re c o d e 17
SAS Hig h-Pe rfo rma nc e Ana lytic s • Ana lyze 100% o f da ta • Mo re / Ne w va ria b le s • Mo re mo de l ite ra tio ns • Ma na g e c o mple x mo de ls • Mo re mo de ls (pe r do ma in a re a ) • Mo re q ue stio ns/ ide a s/ sc e na rio s to e va lua te • Multiple de plo yme nt o ptio ns: b a tc h, re a l-time • Co ntinuo usly mo nito r mo de l e ffe c tive ne ss a nd re tra in 18
SAS Hig h-Pe rfo rma nc e Ana lytic s Pro c e dure s SAS High- SAS High- SAS High- SAS High- SAS High- SAS High- Performance Performance Performance Performance Performance Performance Statistics Econometrics Optimization Data Mining Text Mining Forecasting HPCANDISC HPCDM OPTLSO HPBNET HPTMINE HPFORECAST HPFMM HPCOPULA Select features HPCLUS HPTMSCORE HPPLS HPPSNEL in HPSVM HPPRINCOMP HPCOUNTREG OPTGRAPH HPTSDR HPQUANTSELECT HPSEVERITY OPTMILP HPREDUCE HPLOGISTIC HPQLIM OPTLP HPNEURAL HPREG OPTMODEL HPFOREST HPLMIXED HP4SCORE HPNLIN HPDECIDE HPSPLIT HPGENSELECT The above offers contain these standard PROCS: HPDS2, HPDMDB, HPSAMPLE, HPSUMMARY, HPIMPUTE, HPBIN, HPCORR 19
HPA E xa mple SAS Ana lyst’ s De skto p T ra ditio na l SAS SAS HPA Se rve r E nviro nme nt 20
Pla tfo rms fo r Optima l SAS Pe rfo rma nc e Teradata Active Teradata Appliance Teradata Appliance Data Warehouses for SAS for Hadoop Pro vide s a c o st E na b le s SAS in- Suppo rts SAS in-me mo ry da ta b a se a na lytic s to a na lytic s to o ls use d to e ffe c tive , c le a n a nd pre pa re visua lize da ta a nd b uild pre c o nfig ure d da ta da ta a s we ll a s de plo y da ta mo de ls, g iving sto ra g e a nd sta g ing a nd sc o re mo de ls the m dire c t a c c e ss to a re a tha t ma ke s da ta a va ila b le fo r SAS witho ut ha ving to mo ve da ta within the Ac tive a na lytic s the da ta . Da ta Wa re ho use 21
A F ully Co nta ine d SAS E c o syste m in One Bo x T E RADAT A Data Teradata SAS Teradata 2800 Persistent Appliance Managed Appliance with for SAS Server from SAS Analytics Teradata A fully c ontain SAS Analytic s E c osyste m in a single box fr om T e r adata that inte gr ate s the SAS se r ve r dir e c tly into the c abine t with your data 22
Recommend
More recommend
