ADS Capstone Chronicles Revised

22

Wa bhoevne , i int tiesrkper ye ttion gr et chael lr tehs autl tRs Mi nSTEaibs lien 5t .h1e ss ahmo we su, noint saavse rt ha ge et,ahr og ewt fvaarr oi af bf lteh, es omtohdeeRl ’ sM S E pe xr ea dmi cptl ieo, na sn aRr Me fSrEo mo f t0h. 0e 0o0b1s 2e r3v8e md ev aa nl usetsh. aFto, r of rno ma v tehr ea gaec, t tuhaal tomb soedrevle’ sd psrteodmi catci oh ncsa nd ceevri a t e mt h oi sr tnaul imt ybreart seos ubnydasbvoeurty 0s. m0 0a0l l1, 2i t3i 8s . i mA l pt hoor ut agnht tmo orretcaal il tl yt hraatt et hoeb ms eer avne ds tionmo ua cr he nc at inrceedr a t a s e t (s pp rl iitosr) two adsr 0o .p0p0i0n0g 9a6n3y3r, oswo se voer nt hvee rt yr asi mn -at el ls t Rn oMnSeEos f aorue rqmu iot ed ei ml s ppaecrtffourlm. Geidv evne rt yh awt ec lol .n t e x t , Od rvaesrtai cl la, lml yobdeetlwpeeer nf otrhme atnwcoe md ioddneol itncgh a n g e ampopdreolai nc hg easpupsreoda .cAh de sd, i tt hi oenLailgl yh, tf oGrr abdoitehn t Br aot oe sot if n0g. 1M, ma cahxi inme u( mL i gdhet pGtBhMo )f 2w0i t, ha na dl enaur mn i bn egr oWf hees nt i mt raati onresd eoqnuTa lr at oi n2i n0gpDe raftoarsme te1d abneds tt. e s t e d oA np pTreosat cDha1t a) ,s teht e1 r(ebsoutl ht i nf rgo RmMMS oE dwe lai sn g 02 . 0a n0d0 0t e3s5t9e.dWo hn eTnetsrtaDi naet adsoent 2T r( ba iontihn gf r Do ma t a s e t Mw aosd e0l. i0n0g0A0 p3 p5 r5o, awc hh i c2h) , wt haes rj eu ss ut lstliinggh tRl yMbSeEt t e r than the resulting from Approach 1. 5.2 Model Performance: Adjusted R 2 Ionp taidmdai tl imo no dt oe lc formo mp abr iont gh tohf eoRu Mr aSpEpor fo tahc eh e s , we also evaluated the adjusted R 2 metric of bs toatnhdoapr dt i,mniozne-da dL ij guhs tteGdBRM m o d e l s . T h e 2 metric of a model et hx ep lvaai nr isa bh iol wi t ywi ne ltl hae mt aordgeelt’ sv faerai at ub rl ee s( Ce hx pu lgahi n, 2n0on2-0a)d. jAulsthteoduRgh this is useful, one issue is the 2 always increases or remains tmhoe dsealmi necarse at hs eesnbuemc abuesreoaf df edai nt ugrme soirne tfheea t u r e s

we xipl l l anienveedr idnetchr ee at sa er gt ehte vaamr ioa ub nl et bo yf vt ha roisaet i o n features (Chugh, 2020). Thus, the standard R 2 ma dej ut rdi ci ciast ne oi ft aadgdoi ot ido noanlef ebayt uwrheisc ha rteo u s e f u l i n a mInocdoenltorarsntotto. the standard R 2 metric, the adjusted R 2 metric takes into account the nf ourmmbuel ar ot of fceaal ct uu rl aetset thhaet aa dmj uosdt ee dl hRa s . T h e 2 is included below (Equation 2) where n is the number of observations in the test dataset and k is the number of features in the dataset. (2) Equation Image Obtained From: Chugh, 2020 Because k is in the denominator, there is a pt heenya ldt oy nf oort asdi gdni ni f gi c aa dn dt liyt i iomn pa lr of evaet ut hr ee sRi n i f 2 value (Chugh, 2020). Thus, the adjusted R 2 metric is amnoodpe tl si mwai lt hc haodi ci ef f ef or er nc to nmupma rbienrg owf hf ei ac ht uor fe st wi so a bGei vt teenr tpheartf oo rumr ti nwgommooddeel l. i n g a p p r o a c h e s h a v e af esaitgunrief isc (a5n8t df ei fafteurreensc feoirnMt hoed enlui nmgbAepr porf o a c h 1 vi teirssuusse1f u3l3t of eua tsuertehsef oa dr jMu sotdeedl iRn g A p p r o a c h 2 ) , 2 metric in aodf dt hi tei oLni gthot GRBMMS Emt oo dceolms fpraorme tbhoet hp e r f o r m a n c e approaches. The adjusted R 2 metric for the L0i.9gh5t9G5B. TMhme aoddjeulsftreodmRModeling Approach 1 is 2 metric for the L0 i. 9g h6 t0G7B. IMt ims goodoedl ftroo ms e eMtohdaet lti hn eg aAdpj pu rs ot eadc hR 2 i s 2 mh i ge threi cr bf oerc aMuosde ei lti nr eg aAf fpi rpmr os atchhe 2v ai sl u, ien i fna c t ,

113