%PDF-1.2 % 4 0 obj << /Encoding /WinAnsiEncoding /Subtype /TrueType /Name /F3 /Type /Font /BaseFont /Arial,Bold >> endobj 5 0 obj << /Encoding /WinAnsiEncoding /Subtype /TrueType /Name /F2 /Type /Font /BaseFont /Arial >> endobj 7 0 obj << /Descent 214 /StemV 92 /FontName /Optimum /Ascent 957 /ItalicAngle 0 /Leading 171 /AvgWidth 504 /StemH 92 /FontBBox [-250 -214 1498 957] /Type /FontDescriptor /CapHeight 957 /Flags 32 /MissingWidth 752 /MaxWidth 1248 /XHeight 669 >> endobj 6 0 obj << /Encoding /WinAnsiEncoding /Widths [750 277 285 288 553 553 863 677 161 302 302 409 553 231 221 231 386 553 553 553 553 553 553 553 553 553 553 258 258 500 553 500 379 736 660 585 682 761 480 476 775 759 264 299 604 468 912 760 848 556 849 603 507 512 752 650 991 570 587 649 294 386 294 500 500 500 491 562 487 555 517 276 516 534 229 230 481 220 823 541 572 560 556 326 393 298 543 479 757 466 479 490 295 500 295 500 750 750 750 250 553 421 694 500 500 500 1251 507 243 1027 750 750 750 750 250 250 421 421 600 553 667 500 854 393 243 903 750 750 587 250 258 553 553 553 553 500 500 500 765 338 384 500 750 765 750 357 553 365 365 500 518 553 231 500 365 373 384 865 865 865 352 660 660 660 660 660 660 817 677 480 480 480 480 264 264 264 264 761 760 848 848 848 848 848 553 851 752 752 752 752 587 556 557 491 491 491 491 491 491 793 487 517 517 517 517 229 229 229 229 572 541 572 572 572 572 572 553 596 543 543 543 543 479 560 479] /Subtype /TrueType /Name /F1 /Type /Font /FirstChar 31 /LastChar 255 /FontDescriptor 7 0 R /BaseFont /Optimum >> endobj 9 0 obj << /Descent 220 /StemV 159 /FontName /Optimum,Bold /Ascent 920 /ItalicAngle 0 /Leading 140 /AvgWidth 500 /StemH 159 /FontBBox [-250 -220 1524 920] /Type /FontDescriptor /CapHeight 920 /Flags 16416 /MissingWidth 750 /MaxWidth 1270 /XHeight 644 >> endobj 8 0 obj << /Encoding /WinAnsiEncoding /Widths [750 277 312 387 553 553 863 739 215 344 344 449 553 256 247 256 391 553 553 553 553 553 553 553 553 553 553 256 256 500 553 500 426 771 655 598 677 766 509 493 769 767 324 336 614 488 931 761 849 569 848 615 516 511 754 627 958 583 571 621 347 391 347 500 500 500 510 564 474 565 507 313 513 563 271 278 500 265 847 565 567 565 567 368 399 324 570 471 749 455 469 485 347 500 347 500 750 750 750 295 553 500 768 529 529 500 1266 516 248 1035 750 750 750 750 295 295 500 500 600 553 714 500 865 399 248 838 750 750 571 250 285 553 553 553 553 500 553 500 781 349 394 500 750 781 750 399 553 365 365 500 575 553 256 500 365 366 394 862 862 862 399 655 655 655 655 655 655 875 678 509 509 509 509 324 324 324 324 767 761 849 849 849 849 849 553 848 754 754 754 754 571 569 570 510 510 510 510 510 510 756 474 507 507 507 507 271 271 271 271 567 565 567 567 567 567 567 553 568 570 570 570 570 469 565 469] /Subtype /TrueType /Name /F0 /Type /Font /FirstChar 31 /LastChar 255 /FontDescriptor 9 0 R /BaseFont /Optimum,Bold >> endobj 10 0 obj [/PDF /Text /ImageC] endobj 11 0 obj << /Subtype /Image /Name /im1 /Type /XObject /Width 411 /Filter /DCTDecode /Height 200 /Length 17212 /BitsPerComponent 8 /ColorSpace /DeviceRGB >> stream AdobedC  "$43) )02?@<==8pAS7K}$3,YQ_ֻ!`7?ʶ匆QnHuTeT,=1ʨYK{U"KpI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$'Ve?K[Ve?Kg/kIsserr:tп`'(H%$ "whL=?ʴz?U^unlV3÷p0&L1uFZ!wؙTꌱB鹗e##)pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$U:kppm'3[m+׾s$KPRHAX<"%&-R$rj %:tT.O٬?͓k:x&gwWjU/pI$I$pI$I$pM_4r SwW$@9`MgĪ߶:_JKMkĪl_ޛ__j:]y s`y]m]Sck!sWdG51<yVմdI$_pI$YΕU i *ΕUcAUps-0FG#?]7$}/ѿHD닣đȗL?Kqt??Kqt??K+xYs߽0L[cf0F4l|2Z]S`W=I>Av{X95|֗;@~qȳN?wO?wSD뛣j&\y/?oSD?G?M?"o?i0N̼SM 6q:N̼SM 68y.=}YI$I%pI3 &Ԥ &Ԕ])"^`]ny=o%?[zW'o۸?/,7^z?ʗ/]3m/-/۸_gJ3c@qq?,0\|{fדڹIWqGOҫDTW\t<ޯ?HPW:}SD~/ԿU%%7ԟ?*QWȶ?`Oӫ?OھDwPm_y"C?Tz@,/{@T@,6=\ etܼffwTXHTRtIB5&0AAr5'+>.o4^r#B+ыU ~=pI$I$pI$I$w/1?G1?G:"*3 :$+ѩcj L? 25kI$I%pIguޢޟ`,ԛӰ^HkU9c~+ϋ\u'RW.&IԕLIu_ĝ$"L䥤hDh.T}]ԯY-﫿ɭȻ,.#=w^dy^dNoeLS Gc?f"Of-N ;.wkI$I%pIb}kd54owXZ:yGW}ߗu’BW4J*`%$Usa;6_!\oͪqwV11ݓʇO&MLAzE5[5<ƱkD pI$I$pI3״ZD{{H"= A+ҙjU<p?-?-Z h=溾q/+#Wdiǟ2A5HuN ]֩|OS L_5kTI$KpIpYsk3]Ch=&gmu{uaW-gXP%aJ$h蔦)RP??]ԟy-﫿?_Ȼ,.#=w^b L\xM:)jOf"_f"~Eޮt$I$pt"@KP9Bǃ5π^w׺<|zQsouYI_0S PO?~[h e;XxY<]@Kp7E-UpI$I$pI$\f48p.sZpht{,MeƢ X Iǚq旚xO ஷ?y:~g󭿫_C#_pI$I$pI$I$w//>7~U?ʸ'BBPJ ΋ӥ$uO7KO|O7wz*I$Ip 5o~u!ZU.1qHHlv+$κ?$κ^`n%X(:t$pơ av_RȻ/??[W]? _Ȼ,.#ȼ+̌JBdИ a)OKOf"_f"σ"WxTI$_p]-w+GQ,Z#U|R֟};9\ο%%Q%Yoͮ~qM~neTOT:襔Rʘ!&E,1cDupI$I$pI$\W^FS1;3?\W^FS1;3?\_eLHX*dꗂ_DN&e#/5}쏸/U}쏐lR;WpI$I$pI$I$ </.["'E?q?E @t ''pj8_Чcj@|O0t1TI$K֗jJCZ\L+$I0uC=7X^{ X`>rO/yE ߂=aU RK'XD)DQ柒@}JG?"쾤#Lo}]Mo/_~Ef$?ȼܯ1w%q4QMZǕԯUoY!ڟg5fH->V|{1x]߭OtM?7 zoz_i߼&E[ޥ>^(p]sZuޠzPƙc}{e2Y Y_js߀Y7EP)%!`Iv_Szg?~Sπ]m: _~t6QLz5܊)RƲx@+77B=GsW M_߅?3iWuI?bN%KĪ ׉>oڝ7%UMOĪ }}ktW/ڝ7%UK혟"?kWugK/aĊ Γ[w2 :JYFOɽy }lcYV1a4ɐlpfu<ȹg8ē$+sLdWX sSM)7TT^q>Ҩq:E8Y7;u]_K50$nwpI$I$pI$I$w3WOg\[6?GBQKDէ?ՌO}X?Mz"EդI$Ԇ6bˇnHb},v~7M;7\1+ӄI/:R$IG?"#Lo}\Mw/_~Ef$";/0<0<9@ǔĤX!s[ܤ^|R'N5N|R>)rqKqLaVTЙ)R$.?p$o̪rj|uӱU ؔ?'!7CjV `^MU6.¶65ֈpXzۺ5π\.4l3/ղ՘5c=,3eTe`D J[<'r OJL%.uKR&SvLX Iw ([V:ak&NylY;=p;|ѱ>Ӕ:.w˨_pI$I$pI$I$wGחd#?zC*0!|Rҥ_ R?C z*UդJZju0ւI>FYMOI'&{90ւI_gYq<4x5gu|u .?GY_k}h ԓQ:ςSS4BX*I@HJZ~#Lv_Rȷ&3_~E#";//1#2ꡃW'w(?"o.1>誼lvVk+l쨪lfVѵvʫkX |ޠsc}wc},eo5Y*)$ԯjI9$K$XOD  % ХҺ.gPxK\Π񱅕{ Op nwy从m*yq]Jx6 qOm S 3)a&ԣ#/1ΗcMz2e!eWek,q `u hרeٙmԽ^fn]RLxːʽܸx+ .7)Lr'W5O~ɨ|GX,{1ouVFVfN=#sy*Ծ vFPBQs)pH$d$5NXLkp$tkNާ[cXZaӾ`ӫcX]>՞s ܺ|?=3;lcl*`\;ְִ:I$I$pI$I$pI$I$]yvN?G/1k*,>HT-GODz+ѥXX ?z*'˩ u'˩ uis.PQaXuM:bٙU 'w(feC9y(&R^b|>~S4k>ϭt3 cV]tkfkWX*PjO0SخG+?vR'Y̷%3s.;nMo9y\`:Ħ2E2RկO^WߑWUwZ= RȽzޢ1}]΢1p}> z_;p_ $7tDžcP^Fʨo/<蘸Ȯ1f= ^C1eLꭵV!p ϣtV;=hWi; F`?ʺ/so^G(]?G(X30~BX*ZS)P?p??vR8y̺nL`s_ȻW.'=w^`W;x$RNTJeүLvi;*Tժ~C5~j85UJ󎹞s~`;X<󞹞eAr}K+Ynx#߀Y9U`'5KDBa7eox|;L+~kY>p%"9k E)(Hj=d5!tSs<=tSs<=lt ;77pI$I$pI$I$pI$I$]yvGHU*c*zH ӣ|W||o髽,PKGwTF@Ŭ)>7~u?d7µ=[;w \K~>) >+0 $a\.Ϯf\|[.RqhPwIU[X heUVګk!hkD ~VO2 11$E <BK#3gG`k?/sȻw.#=w^ayq1HrrCj"Pv;-.Ww]־lEvvGaݦu2 _Mnp\ZJB%! XS^qU,G~f]T0jd_]M>K1f=  e2k.ƺUmcD5DER_>~yo.~}ev?H ίqG.~Pt2Ʈ:ܬnnS(j*6;/mLS=Z i(OX8IX>9ė8>2^9&O&B$5O XO#t_{.N._Kq= _Un[yi{_hs]G+/W,䟞%" Bu}GE˲#3As-s_ȻW.#=w^^H\be얝_ԭYoW@V淬V`Z=Qak"[9[>vys`;X<R^[!P* ΩR! .wM[5wuSnߘ񫽬w+w;%W{YWPI/U?!Z]G+ɮB>!7d |׽'r/ !wW7fEd4x{Cm֞d4~U.m$I&:I[4oWu,۠0汕^*RLx'<%Ֆ֟sǂ&5/Bh .ht.Whs]?G+/W$Jy7o4']ԏy̻Ȳ??tVs_ȻW.#=w^^%y{\bA &!+ªX`+_KaAlF=[ TFUI)"u_~f]T0jd3.5{;윊o/1 qf= `/IơWS5.ƪUlEERI$?!Z]G+ 玅/4xNԳ35Wor}Mk-7ϫ koǺ\I% leU90ւIPUSyJg810WNeqN~Gs.'ddYoH*P)p0HcĐ*cF@Rh$nʘnʘݭBtI$_pI$I$pI$I$]y~F?G/H1=(H]y1;H_+uRSwLuO};'&TE辭 Ӿ۸~s.b}wԒIs]?G+?!YX0~B… PT@2T@2 Rx+#0~E}HG?" s_ȻW.#=w^`bWW06!(n`pqG(.B\ܔkvVǍ]gùZSnߘ񫽬o}^djk>PI$I~✞{Z%y.)7 sFKԛ-hZ7^xrSr` BmOĦ)0sH ZnYc0-o{װÚAu3m61|}E߭oO'\܁Zw}`UVx ԭ;Jn7RY԰k`qx Ԯ[}b7ѩ{뫾+X ii]UCӬWXRrZOybHZý>bHZýt ýpI$I$pI$I$pI$I$]ynGr?\](r*2)#UFICTnVnrqgO9y{OOeq~і,|'pI$\O2 VgLLX*DI"JI]ԏyv_R?dL~Eo5es~E#Oo;Ɓ2I$_!`HSR'WWC|,WC&=T\ahf5 `.ccK5Zkm52kM$I$x+piWΜA?^w:s~{VsarGe9:VtORRuLiM)Je)MqC&V 2LɂZ!jNޖlpX1Z(97L@ޛ-rи~?"W?ף OGyG&^Vc:]Os };[ٙ _uںV]2ݻvZY]_pIWͣmN_;6 8rjŨaӀ$-jk0<Yͻ1T _*mŬ^%r'Fk]H?@PhfMv{ ~}]kџDv@ͤ^~ 7Psdh?EI/ <//hdZ?*.6ʅ n1*03L8L#J\"e;GhįD s%u;b1uwįpI$X?\-r?!YX?1‘$)袢j! Se#.O]ԏ?o\"ȿAߑvY]~EM?y{Jr}0Kㅃ‘N-Bx:}W[7|;}M?1Sdw[Wap[pI$I$p]}Kcua"}.XtU8M,Շr1ǹX4y[si _]XhAC(se<qI$%~$X z%)> ӤR X>jpI#cdY{.^!"kunme0 ORƑ+кWQ@{t.}G7pIa>nUӊ]@rzX)ӈ]M=ܬve+V?ܢF+4Z1r]#ie`4:)ӓs~?ʿZ uN-.6N%m-c>4Y[: N۲ Amx^Kv/cgEIuQq4=ދ>,rߙ[1Br]ǿk#y^VW <.yXj7~U9cWp5NAl$t0<@_ā[Tp7u{NYN{i,AnAW]WDpI$X?\ԦV/)f}`BBX*dPxK? qHɲ?">&wL~E~E%?"5^\F@M KDDXJdjvGƳ+*Y|KKQ~NM2i |}]WЛiڴ$ǿ㍵m4ǿ㍵m4e6DW<#ESH<>oQ#UW=W ] w^c[nPG(#TTt[|S 4rL4hQHMn c{X\5~*o-{GLª>'pۍʇO+HpI$X_\ԡV/$7\7+LxnvM$қ>'+c-﫟![~E&D/tL3Lה)KIXJd'H;G=d+1 '[Wt~CY]R_pI$I$pI$QX״9BƖ;ep2 T2A;mciUqK0YWTІ=_f+-ԫ .gR߫böV1Aaa[{i !-pA0 n#( /X0M8$J~wQ3+og7~~w,?޹iNk=+G4!tǥukq#-8wpzWW2_Sgw 3Cq}gF?Tfcz{DYS] ;([Vu=Z܊VaE_Ρ{%g_Ρ t.![*RfӼAs:|{:'HyG}XwX;oJ鬭uV>+l2Htl7`2˟#?Nv.#+qs'u]VWpUY黷L]qzh~* 6wT6?rqM8wܜ~v;YO;F~IO6}'Z]'._0??_TGTO.O_cʘIk L-|z[K*a%7+FUMqFu#FuD#VUFp ʇ G%f7`)i%R)ܪ,0vG;k1k?O[p-f%x7A ƮtDl2299 p˸v.;]~i5u$_pGIm moI@KhI x.h]II$I%pI$I$ .p$PI$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$ endstream endobj 12 0 obj << /Length 666 >> stream 0.72 w 1 J 1 j 0.2 0.2 0.6 RG 1 1 1 rg 72 692.88 468.12 41.4 re b BT 72 727.08 TD 0 0 0 rg /F0 12 Tf 0 Tc 3.276 Tw ( ) Tj 0 -16.08 TD 0 0 0.502 rg /F1 14.04 Tf 0 Tc -0.0494 Tw ( Numerical Algorithms Group) Tj 176.4 0 TD 0 Tc 3.9109 Tw ( ) Tj -176.4 -13.2 TD 0 0 0 rg /F2 11.04 Tf 2.8109 Tw ( ) Tj 0 -12.6 TD ( ) Tj 0 -12.72 TD ( ) Tj 0 -634.2 TD /F2 9.96 Tf -0.0205 Tc 0 Tw (\251) Tj 7.32 0 TD 0.0259 Tc -0.0348 Tw ( The Numerical Algorithms Group) Tj 148.44 0 TD 0 Tc 57.4711 Tw ( ) Tj 60.24 0 TD 237.7111 Tw ( ) Tj 240.48 0 TD -0.0178 Tc 0 Tw (1) Tj 5.52 0 TD 0 Tc 2.6311 Tw ( ) Tj ET q 72 0 0 -35.28 462 729 cm /im1 Do endstream endobj 13 0 obj << /Length 7371 >> stream Q BT 72 659.88 TD /F2 11.04 Tf 0.0127 Tc 0 Tw (Title:) Tj 24 0 TD 0 Tc 3.0509 Tw ( ) Tj 41.4 0.12 TD /F3 11.04 Tf -0.0081 Tc 0.059 Tw (Examples of the Use of Data Mining in Financial Applications) Tj 320.28 0 TD 0 Tc 3.0509 Tw ( ) Tj -385.68 -12.84 TD /F2 11.04 Tf ( ) Tj 65.4 0 TD ( ) Tj -65.4 -12.6 TD -0.0023 Tc 0 Tw (Summary:) Tj 50.28 0 TD 0 Tc 3.0509 Tw ( ) Tj 15.12 0.96 TD /F2 9.96 Tf -0.0039 Tc -0.005 Tw (This article considers building mathematical models with financial data by using data ) Tj 0 -11.52 TD 0.0019 Tc -0.0108 Tw (mining techniques. In general, data mining methods such as neural networks and decision ) Tj T* 0.011 Tc -0.0199 Tw (trees can be a usef) Tj 84.84 0 TD -0.021 Tc 0.0121 Tw (ul addition to the techniques available to the financial analyst. ) Tj -84.84 -11.52 TD -0.0049 Tc -0.004 Tw (However, the data mining techniques tend to require more historical data than the standard ) Tj 0 -11.4 TD 0.003 Tc -0.0119 Tw (models and, in the case of neural networks, can be difficult to interpret.) Tj 311.76 0 TD 0 Tc 2.7511 Tw ( ) Tj -377.16 -21.36 TD 2.6311 Tw ( ) Tj 0 -11.4 TD -0.0031 Tc 0.8526 Tw (This article considers building mathematical models with financial data by using data mining techniques.) Tj 0 Tc -0.0089 Tw ( ) Tj 0 -11.52 TD 0.004 Tc 0.7631 Tw (In general, data mining methods such as neural networks and decision trees can be a usef) Tj 410.4 0 TD -0.0246 Tc 0.7357 Tw (ul addition to) Tj 0 Tc -0.0089 Tw ( ) Tj -410.4 -11.52 TD -0.0144 Tc 1.4776 Tw (the techniques available to the financial analyst. However, the data mining techniques tend to require) Tj 0 Tc -0.0089 Tw ( ) Tj 0 -11.52 TD 0.0031 Tc 1.6539 Tw (more historical data than the standard models and, in the case of neural networks, can be difficult to) Tj 0 Tc -0.0089 Tw ( ) Tj T* -0.0062 Tc 0 Tw (interpret.) Tj 39.24 0 TD 0 Tc 2.6311 Tw ( ) Tj -39.24 -11.4 TD ( ) Tj 0 -11.52 TD 0.0336 Tc -0.0425 Tw (Stock market returns a) Tj 100.8 0 TD -0.0035 Tc -0.0054 Tw (nd foreign currency exchange rates data can be thought to fall into one of four ) Tj -100.8 -11.52 TD -0.0015 Tc -0.0074 Tw (categories as follows.) Tj 95.16 0 TD 0 Tc 2.6311 Tw ( ) Tj -95.16 -11.52 TD ( ) Tj 0 -14.52 TD -0.0133 Tc 0 Tw (1.) Tj 8.28 0 TD 0 Tc 6.9511 Tw ( ) Tj 9.72 0 TD -0.0125 Tc 0.0036 Tw (Five time series: index value at open, index value at close, highest index value, lowest index value ) Tj 0 -11.52 TD -0.0054 Tc -0.0035 Tw (and trading volume.) Tj 87.36 0 TD 0 Tc 2.6311 Tw ( ) Tj -105.36 -14.4 TD -0.0133 Tc 0 Tw (2.) Tj 8.28 0 TD 0 Tc 6.9511 Tw ( ) Tj 9.72 0 TD 0.0079 Tc -0.0168 Tw (Fundamental factors: e.g., th) Tj 129.72 0 TD -0.0054 Tc -0.0035 Tw (e price of gold, retail sales index, industrial production indices, foreign ) Tj -129.72 -11.52 TD -0.0038 Tc -0.0051 Tw (currency exchange rates.) Tj 111.72 0 TD 0 Tc 2.6311 Tw ( ) Tj -129.72 -14.52 TD -0.0133 Tc 0 Tw (3.) Tj 8.28 0 TD 0 Tc 6.9511 Tw ( ) Tj 9.72 0 TD 0.0142 Tc -0.0231 Tw (Lagged returns from the time series of interest.) Tj 207 0 TD 0 Tc 2.6311 Tw ( ) Tj -225 -14.52 TD -0.0133 Tc 0 Tw (4.) Tj 8.28 0 TD 0 Tc 6.9511 Tw ( ) Tj 9.72 0 TD 0.0082 Tc -0.0171 Tw (Technical factors: variables that are functions of one more time series, e.g., moving averages.) Tj 416.28 0 TD 0 Tc 2.6311 Tw ( ) Tj -434.28 -11.52 TD ( ) Tj 0 -11.52 TD 0.0192 Tc -0.0281 Tw (The stan) Tj 38.88 0 TD 0.009 Tc -0.0178 Tw (dard approach to modeling stock market returns or exchange rates is to model the univariate ) Tj -38.88 -11.4 TD 0.0029 Tc -0.0118 Tw (time series with autoregressive \(AR\) and moving average \(MA\) models. A trader can determine an ) Tj 0 -11.52 TD 0.0048 Tc -0.0137 Tw (appropriate number of lags for AR and ARMA models based on experienc) Tj 325.68 0 TD -0.0436 Tc 0.0347 Tw (e and by analyzing the time ) Tj -325.68 -11.52 TD 0.0063 Tc -0.0152 Tw (series data. Similarly, an appropriate number of regimes for SETAR \(self) Tj 322.44 0 TD 0.0433 Tc 0 Tw (-) Tj 3.36 0 TD -0.0074 Tc -0.0015 Tw (exciting transition AR\) and ) Tj -325.8 -11.52 TD 0.0117 Tc -0.0205 Tw (STAR \(smooth transition AR\) models can be determined. These models are deterministic in the sense ) Tj 0 -11.52 TD -0.0023 Tc -0.0066 Tw (that they attempt to use ma) Tj 120.6 0 TD 0.0035 Tc -0.0124 Tw (thematical equations to describe the process that generates the time series. ) Tj -120.6 -11.4 TD -0.0103 Tc 0.0014 Tw (The advantage of these models lies in their interpretability.) Tj 256.8 0 TD 0 Tc 2.6311 Tw ( ) Tj -256.8 -11.52 TD ( ) Tj 0 -11.52 TD -0.0008 Tc -0.0081 Tw (Another approach, drawn from data mining, is to adopt a model that is flexible in the sense that it can ) Tj T* 0.0146 Tc 0 Tw (approximate) Tj 54.96 0 TD -0.0042 Tc -0.0046 Tw ( a wide class of functions with high accuracy. Such models are non) Tj 297.48 0 TD 0.0433 Tc 0 Tw (-) Tj 3.36 0 TD 0.0096 Tc -0.0185 Tw (parametric in the sense ) Tj -355.8 -11.52 TD -0.0059 Tc -0.003 Tw (that there need not be a direct relationship between the parameter values of a fitted model and the data. ) Tj 0 -11.52 TD 0.0026 Tc -0.0115 Tw (The advantages of using such a model include:) Tj 208.2 0 TD 0 Tc 2.6311 Tw ( ) Tj -208.2 -11.4 TD ( ) Tj 0 -11.52 TD -0.0133 Tc 0 Tw (1.) Tj 8.28 0 TD 0 Tc 6.9511 Tw ( ) Tj 9.72 0 TD -0.0088 Tc -0 Tw (The abilit) Tj 40.32 0 TD -0.0107 Tc 0.0019 Tw (y to model highly complex functions.) Tj 159.6 0 TD 0 Tc 2.6311 Tw ( ) Tj -217.92 -14.52 TD -0.0133 Tc 0 Tw (2.) Tj 8.28 0 TD 0 Tc 6.9511 Tw ( ) Tj 9.72 0 TD -0.0068 Tc -0.0021 Tw (The ability to use a high number of variables in the model and, therefore, to include other data \(i.e. ) Tj 0 -11.52 TD 0 Tc -0.0093 Tw (fundamental and technical factors\) in addition to lagged time series data.) Tj 319.92 0 TD 0 Tc 2.6311 Tw ( ) Tj -337.92 -14.52 TD ( ) Tj 0 -11.52 TD -0.0022 Tc -0.0067 Tw (The disadvantage of non) Tj 109.56 0 TD 0.0433 Tc 0 Tw (-) Tj 3.36 0 TD 0.0239 Tc -0.0327 Tw (parametric model) Tj 77.28 0 TD -0.0221 Tc 0.0132 Tw (s is that they are not easy to interpret.) Tj 165.84 0 TD 0 Tc 2.6311 Tw ( ) Tj -356.04 -11.4 TD ( ) Tj 0 -11.52 TD -0 Tc -0.0088 Tw (In the case of data mining time series data, the model of choice is a neural network. By adjusting the ) Tj T* 0.0065 Tc -0.0154 Tw (number of free parameters associated with a model, a trader controls its flexibility. Often, cross) Tj 420 0 TD 0.0433 Tc 0 Tw (-) Tj -420 -11.52 TD -0.0326 Tc (validatio) Tj 36.24 0 TD -0.0132 Tc 0.0043 Tw (n, or hold) Tj 41.4 0 TD 0.0433 Tc 0 Tw (-) Tj 3.36 0 TD 0.0069 Tc -0.0157 Tw (out data, is used to determine a suitable value for the number of free parameters ) Tj -81 -11.52 TD 0.0031 Tc -0.0119 Tw (contained in a neural network structure. The neural network most commonly used in financial ) Tj 0 -11.52 TD -0.0032 Tc -0.0057 Tw (applications is a multi) Tj 94.56 0 TD 0.0433 Tc 0 Tw (-) Tj 3.36 0 TD -0.0209 Tc 0.012 Tw (layer perceptron \(MLP\) with a single hidden la) Tj 201.72 0 TD -0.0183 Tc 0.0094 Tw (yer of nodes. ) Tj 60.12 0 TD 0 Tc 2.6311 Tw ( ) Tj -359.76 -11.4 TD ( ) Tj ET 0 0 0 RG 72 579.24 m 540 579.24 l S endstream endobj 3 0 obj << /Resources << /Font << /F3 4 0 R /F2 5 0 R /F1 6 0 R /F0 8 0 R >> /ProcSet 10 0 R /XObject << /im1 11 0 R >> >> /Type /Page /Parent 1 0 R /Contents [12 0 R 13 0 R] /MediaBox [0 0 612 792] >> endobj 16 0 obj << /Encoding /WinAnsiEncoding /Subtype /TrueType /Name /F4 /Type /Font /BaseFont /Arial,Italic >> endobj 17 0 obj << /Subtype /Image /Name /im2 /Type /XObject /Width 411 /Filter /DCTDecode /Height 200 /Length 17212 /BitsPerComponent 8 /ColorSpace /DeviceRGB >> stream AdobedC  "$43) )02?@<==8pAS7K}$3,YQ_ֻ!`7?ʶ匆QnHuTeT,=1ʨYK{U"KpI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$'Ve?K[Ve?Kg/kIsserr:tп`'(H%$ "whL=?ʴz?U^unlV3÷p0&L1uFZ!wؙTꌱB鹗e##)pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$U:kppm'3[m+׾s$KPRHAX<"%&-R$rj %:tT.O٬?͓k:x&gwWjU/pI$I$pI$I$pM_4r SwW$@9`MgĪ߶:_JKMkĪl_ޛ__j:]y s`y]m]Sck!sWdG51<yVմdI$_pI$YΕU i *ΕUcAUps-0FG#?]7$}/ѿHD닣đȗL?Kqt??Kqt??K+xYs߽0L[cf0F4l|2Z]S`W=I>Av{X95|֗;@~qȳN?wO?wSD뛣j&\y/?oSD?G?M?"o?i0N̼SM 6q:N̼SM 68y.=}YI$I%pI3 &Ԥ &Ԕ])"^`]ny=o%?[zW'o۸?/,7^z?ʗ/]3m/-/۸_gJ3c@qq?,0\|{fדڹIWqGOҫDTW\t<ޯ?HPW:}SD~/ԿU%%7ԟ?*QWȶ?`Oӫ?OھDwPm_y"C?Tz@,/{@T@,6=\ etܼffwTXHTRtIB5&0AAr5'+>.o4^r#B+ыU ~=pI$I$pI$I$w/1?G1?G:"*3 :$+ѩcj L? 25kI$I%pIguޢޟ`,ԛӰ^HkU9c~+ϋ\u'RW.&IԕLIu_ĝ$"L䥤hDh.T}]ԯY-﫿ɭȻ,.#=w^dy^dNoeLS Gc?f"Of-N ;.wkI$I%pIb}kd54owXZ:yGW}ߗu’BW4J*`%$Usa;6_!\oͪqwV11ݓʇO&MLAzE5[5<ƱkD pI$I$pI3״ZD{{H"= A+ҙjU<p?-?-Z h=溾q/+#Wdiǟ2A5HuN ]֩|OS L_5kTI$KpIpYsk3]Ch=&gmu{uaW-gXP%aJ$h蔦)RP??]ԟy-﫿?_Ȼ,.#=w^b L\xM:)jOf"_f"~Eޮt$I$pt"@KP9Bǃ5π^w׺<|zQsouYI_0S PO?~[h e;XxY<]@Kp7E-UpI$I$pI$\f48p.sZpht{,MeƢ X Iǚq旚xO ஷ?y:~g󭿫_C#_pI$I$pI$I$w//>7~U?ʸ'BBPJ ΋ӥ$uO7KO|O7wz*I$Ip 5o~u!ZU.1qHHlv+$κ?$κ^`n%X(:t$pơ av_RȻ/??[W]? _Ȼ,.#ȼ+̌JBdИ a)OKOf"_f"σ"WxTI$_p]-w+GQ,Z#U|R֟};9\ο%%Q%Yoͮ~qM~neTOT:襔Rʘ!&E,1cDupI$I$pI$\W^FS1;3?\W^FS1;3?\_eLHX*dꗂ_DN&e#/5}쏸/U}쏐lR;WpI$I$pI$I$ </.["'E?q?E @t ''pj8_Чcj@|O0t1TI$K֗jJCZ\L+$I0uC=7X^{ X`>rO/yE ߂=aU RK'XD)DQ柒@}JG?"쾤#Lo}]Mo/_~Ef$?ȼܯ1w%q4QMZǕԯUoY!ڟg5fH->V|{1x]߭OtM?7 zoz_i߼&E[ޥ>^(p]sZuޠzPƙc}{e2Y Y_js߀Y7EP)%!`Iv_Szg?~Sπ]m: _~t6QLz5܊)RƲx@+77B=GsW M_߅?3iWuI?bN%KĪ ׉>oڝ7%UMOĪ }}ktW/ڝ7%UK혟"?kWugK/aĊ Γ[w2 :JYFOɽy }lcYV1a4ɐlpfu<ȹg8ē$+sLdWX sSM)7TT^q>Ҩq:E8Y7;u]_K50$nwpI$I$pI$I$w3WOg\[6?GBQKDէ?ՌO}X?Mz"EդI$Ԇ6bˇnHb},v~7M;7\1+ӄI/:R$IG?"#Lo}\Mw/_~Ef$";/0<0<9@ǔĤX!s[ܤ^|R'N5N|R>)rqKqLaVTЙ)R$.?p$o̪rj|uӱU ؔ?'!7CjV `^MU6.¶65ֈpXzۺ5π\.4l3/ղ՘5c=,3eTe`D J[<'r OJL%.uKR&SvLX Iw ([V:ak&NylY;=p;|ѱ>Ӕ:.w˨_pI$I$pI$I$wGחd#?zC*0!|Rҥ_ R?C z*UդJZju0ւI>FYMOI'&{90ւI_gYq<4x5gu|u .?GY_k}h ԓQ:ςSS4BX*I@HJZ~#Lv_Rȷ&3_~E#";//1#2ꡃW'w(?"o.1>誼lvVk+l쨪lfVѵvʫkX |ޠsc}wc},eo5Y*)$ԯjI9$K$XOD  % ХҺ.gPxK\Π񱅕{ Op nwy从m*yq]Jx6 qOm S 3)a&ԣ#/1ΗcMz2e!eWek,q `u hרeٙmԽ^fn]RLxːʽܸx+ .7)Lr'W5O~ɨ|GX,{1ouVFVfN=#sy*Ծ vFPBQs)pH$d$5NXLkp$tkNާ[cXZaӾ`ӫcX]>՞s ܺ|?=3;lcl*`\;ְִ:I$I$pI$I$pI$I$]yvN?G/1k*,>HT-GODz+ѥXX ?z*'˩ u'˩ uis.PQaXuM:bٙU 'w(feC9y(&R^b|>~S4k>ϭt3 cV]tkfkWX*PjO0SخG+?vR'Y̷%3s.;nMo9y\`:Ħ2E2RկO^WߑWUwZ= RȽzޢ1}]΢1p}> z_;p_ $7tDžcP^Fʨo/<蘸Ȯ1f= ^C1eLꭵV!p ϣtV;=hWi; F`?ʺ/so^G(]?G(X30~BX*ZS)P?p??vR8y̺nL`s_ȻW.'=w^`W;x$RNTJeүLvi;*Tժ~C5~j85UJ󎹞s~`;X<󞹞eAr}K+Ynx#߀Y9U`'5KDBa7eox|;L+~kY>p%"9k E)(Hj=d5!tSs<=tSs<=lt ;77pI$I$pI$I$pI$I$]yvGHU*c*zH ӣ|W||o髽,PKGwTF@Ŭ)>7~u?d7µ=[;w \K~>) >+0 $a\.Ϯf\|[.RqhPwIU[X heUVګk!hkD ~VO2 11$E <BK#3gG`k?/sȻw.#=w^ayq1HrrCj"Pv;-.Ww]־lEvvGaݦu2 _Mnp\ZJB%! XS^qU,G~f]T0jd_]M>K1f=  e2k.ƺUmcD5DER_>~yo.~}ev?H ίqG.~Pt2Ʈ:ܬnnS(j*6;/mLS=Z i(OX8IX>9ė8>2^9&O&B$5O XO#t_{.N._Kq= _Un[yi{_hs]G+/W,䟞%" Bu}GE˲#3As-s_ȻW.#=w^^H\be얝_ԭYoW@V淬V`Z=Qak"[9[>vys`;X<R^[!P* ΩR! .wM[5wuSnߘ񫽬w+w;%W{YWPI/U?!Z]G+ɮB>!7d |׽'r/ !wW7fEd4x{Cm֞d4~U.m$I&:I[4oWu,۠0汕^*RLx'<%Ֆ֟sǂ&5/Bh .ht.Whs]?G+/W$Jy7o4']ԏy̻Ȳ??tVs_ȻW.#=w^^%y{\bA &!+ªX`+_KaAlF=[ TFUI)"u_~f]T0jd3.5{;윊o/1 qf= `/IơWS5.ƪUlEERI$?!Z]G+ 玅/4xNԳ35Wor}Mk-7ϫ koǺ\I% leU90ւIPUSyJg810WNeqN~Gs.'ddYoH*P)p0HcĐ*cF@Rh$nʘnʘݭBtI$_pI$I$pI$I$]y~F?G/H1=(H]y1;H_+uRSwLuO};'&TE辭 Ӿ۸~s.b}wԒIs]?G+?!YX0~B… PT@2T@2 Rx+#0~E}HG?" s_ȻW.#=w^`bWW06!(n`pqG(.B\ܔkvVǍ]gùZSnߘ񫽬o}^djk>PI$I~✞{Z%y.)7 sFKԛ-hZ7^xrSr` BmOĦ)0sH ZnYc0-o{װÚAu3m61|}E߭oO'\܁Zw}`UVx ԭ;Jn7RY԰k`qx Ԯ[}b7ѩ{뫾+X ii]UCӬWXRrZOybHZý>bHZýt ýpI$I$pI$I$pI$I$]ynGr?\](r*2)#UFICTnVnrqgO9y{OOeq~і,|'pI$\O2 VgLLX*DI"JI]ԏyv_R?dL~Eo5es~E#Oo;Ɓ2I$_!`HSR'WWC|,WC&=T\ahf5 `.ccK5Zkm52kM$I$x+piWΜA?^w:s~{VsarGe9:VtORRuLiM)Je)MqC&V 2LɂZ!jNޖlpX1Z(97L@ޛ-rи~?"W?ף OGyG&^Vc:]Os };[ٙ _uںV]2ݻvZY]_pIWͣmN_;6 8rjŨaӀ$-jk0<Yͻ1T _*mŬ^%r'Fk]H?@PhfMv{ ~}]kџDv@ͤ^~ 7Psdh?EI/ <//hdZ?*.6ʅ n1*03L8L#J\"e;GhįD s%u;b1uwįpI$X?\-r?!YX?1‘$)袢j! Se#.O]ԏ?o\"ȿAߑvY]~EM?y{Jr}0Kㅃ‘N-Bx:}W[7|;}M?1Sdw[Wap[pI$I$p]}Kcua"}.XtU8M,Շr1ǹX4y[si _]XhAC(se<qI$%~$X z%)> ӤR X>jpI#cdY{.^!"kunme0 ORƑ+кWQ@{t.}G7pIa>nUӊ]@rzX)ӈ]M=ܬve+V?ܢF+4Z1r]#ie`4:)ӓs~?ʿZ uN-.6N%m-c>4Y[: N۲ Amx^Kv/cgEIuQq4=ދ>,rߙ[1Br]ǿk#y^VW <.yXj7~U9cWp5NAl$t0<@_ā[Tp7u{NYN{i,AnAW]WDpI$X?\ԦV/)f}`BBX*dPxK? qHɲ?">&wL~E~E%?"5^\F@M KDDXJdjvGƳ+*Y|KKQ~NM2i |}]WЛiڴ$ǿ㍵m4ǿ㍵m4e6DW<#ESH<>oQ#UW=W ] w^c[nPG(#TTt[|S 4rL4hQHMn c{X\5~*o-{GLª>'pۍʇO+HpI$X_\ԡV/$7\7+LxnvM$қ>'+c-﫟![~E&D/tL3Lה)KIXJd'H;G=d+1 '[Wt~CY]R_pI$I$pI$QX״9BƖ;ep2 T2A;mciUqK0YWTІ=_f+-ԫ .gR߫böV1Aaa[{i !-pA0 n#( /X0M8$J~wQ3+og7~~w,?޹iNk=+G4!tǥukq#-8wpzWW2_Sgw 3Cq}gF?Tfcz{DYS] ;([Vu=Z܊VaE_Ρ{%g_Ρ t.![*RfӼAs:|{:'HyG}XwX;oJ鬭uV>+l2Htl7`2˟#?Nv.#+qs'u]VWpUY黷L]qzh~* 6wT6?rqM8wܜ~v;YO;F~IO6}'Z]'._0??_TGTO.O_cʘIk L-|z[K*a%7+FUMqFu#FuD#VUFp ʇ G%f7`)i%R)ܪ,0vG;k1k?O[p-f%x7A ƮtDl2299 p˸v.;]~i5u$_pGIm moI@KhI x.h]II$I%pI$I$ .p$PI$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$pI$I$ endstream endobj 18 0 obj << /Length 666 >> stream 0.72 w 1 J 1 j 0.2 0.2 0.6 RG 1 1 1 rg 72 692.88 468.12 41.4 re b BT 72 727.08 TD 0 0 0 rg /F0 12 Tf 0 Tc 3.276 Tw ( ) Tj 0 -16.08 TD 0 0 0.502 rg /F1 14.04 Tf 0 Tc -0.0494 Tw ( Numerical Algorithms Group) Tj 176.4 0 TD 0 Tc 3.9109 Tw ( ) Tj -176.4 -13.2 TD 0 0 0 rg /F2 11.04 Tf 2.8109 Tw ( ) Tj 0 -12.6 TD ( ) Tj 0 -12.72 TD ( ) Tj 0 -634.2 TD /F2 9.96 Tf -0.0205 Tc 0 Tw (\251) Tj 7.32 0 TD 0.0259 Tc -0.0348 Tw ( The Numerical Algorithms Group) Tj 148.44 0 TD 0 Tc 57.4711 Tw ( ) Tj 60.24 0 TD 237.7111 Tw ( ) Tj 240.48 0 TD -0.0178 Tc 0 Tw (2) Tj 5.52 0 TD 0 Tc 2.6311 Tw ( ) Tj ET q 72 0 0 -35.28 462 729 cm /im2 Do endstream endobj 19 0 obj << /Length 6795 >> stream Q BT 72 660.96 TD 0.0203 Tc -0.0292 Tw (The problem of predicting stock market returns or exchange rates at time ) Tj 324.96 0 TD /F4 9.96 Tf -0.0278 Tc 0 Tw (t+1) Tj 14.04 0 TD /F2 9.96 Tf 0.002 Tc -0.0109 Tw ( can be cast as either a ) Tj -339 -11.64 TD 0.0087 Tc -0.0175 Tw (regression or classification problem. ) Tj 0.0116 Tc -0.0205 Tw (Whereas the regression problem for exchange rate data involves ) Tj 0 -11.52 TD -0.0028 Tc -0.0061 Tw (modeling the actual exchange rate, th) Tj 165.96 0 TD -0.0051 Tc -0.0038 Tw (e classification problem involves predicting whether the exchange ) Tj -165.96 -11.52 TD 0.0043 Tc -0.0132 Tw (rate has increased or decreased.) Tj 145.68 0 TD 0 Tc 2.6311 Tw ( ) Tj -145.68 -11.52 TD ( ) Tj 0 -11.4 TD 0.0085 Tc -0.0174 Tw (Applications that involve modeling returns from the stock market include portfolio management and ) Tj 0 -11.52 TD -0.0019 Tc -0.0069 Tw (trading futures \(see below\).) Tj 120.6 0 TD 0 Tc 2.6311 Tw ( ) Tj -120.6 -11.52 TD ( ) Tj 0 -12.36 TD /F3 11.04 Tf -0.0032 Tc 0.0541 Tw (MLP regression example: por) Tj 157.8 0 TD 0.0052 Tc 0.0457 Tw (tfolio management) Tj 98.28 0 TD 0 Tc 3.0509 Tw ( ) Tj -256.08 -12.72 TD /F2 11.04 Tf 2.8109 Tw ( ) Tj 0 -11.76 TD /F2 9.96 Tf -0.0032 Tc -0.0057 Tw (The regression case involves predicting the \(raw\) return values. Such predictions can be used to ) Tj 0 -11.52 TD 0.013 Tc -0.0218 Tw (manage a portfolio of n stocks as follows.) Tj 182.52 0 TD 0 Tc 2.6311 Tw ( ) Tj -182.52 -11.4 TD ( ) Tj 0 -11.52 TD 0.0069 Tc -0.0158 Tw (Suppose that historical data for N \(N > n\) stocks are used to fit N multi) Tj 308.88 0 TD 0.0433 Tc 0 Tw (-) Tj 3.36 0 TD -0.0169 Tc 0.008 Tw (layer perceptrons. A) Tj 91.56 0 TD -0.0015 Tc -0.0074 Tw (t the end of ) Tj -403.8 -11.52 TD 0 Tc -0.0094 Tw (each week the MLPs are re) Tj 121.2 0 TD 0.0433 Tc 0 Tw (-) Tj 3.36 0 TD -0.0022 Tc -0.0067 Tw (fitted to include the latest historical data. For example, suppose company Z\222s ) Tj -124.56 -11.52 TD 0.0031 Tc -0.012 Tw (pension fund has been managing a portfolio of $100 million since December 1993 using multi) Tj 410.88 0 TD 0.0433 Tc 0 Tw (-) Tj 3.36 0 TD -0.0687 Tc 0.0598 Tw (layer ) Tj -414.24 -11.52 TD 0.0101 Tc -0.019 Tw (perceptrons. The fund monitors a pool o) Tj 179.64 0 TD 0.009 Tc -0.0178 Tw (f 1,000 U.S. stocks on a weekly basis. For each of these stocks ) Tj -179.64 -11.4 TD 0.0132 Tc -0.0221 Tw (there is a MLP which models the future performance of the stock as a function of the stock\222s exposure to ) Tj 0 -11.52 TD -0 Tc -0.0085 Tw (40 fundamental and technical factors, and gives an estimate of its weekly price change) Tj 381.84 0 TD 0.002 Tc -0.0109 Tw (. The company ) Tj -381.84 -11.52 TD -0.0031 Tc -0.0058 Tw (then selects a portfolio of the top n stocks and allocates the fund proportionately to predicted returns.) Tj 443.04 0 TD 0 Tc 2.6311 Tw ( ) Tj -443.04 -12.48 TD /F2 11.04 Tf 2.8109 Tw ( ) Tj 0 -12.48 TD /F3 11.04 Tf 0.0046 Tc 0.0463 Tw (MLP classification example: trading futures) Tj 231.72 0 TD 0 Tc 3.0509 Tw ( ) Tj -231.72 -12.84 TD /F2 11.04 Tf 2.8109 Tw ( ) Tj 0 -11.64 TD /F2 9.96 Tf 0.0102 Tc -0.0191 Tw (As an example of a data mining classifier, consider the problem of trading a future of s) Tj 379.8 0 TD 0.0054 Tc -0.0143 Tw (tock A at price B on ) Tj -379.8 -11.52 TD -0.0174 Tc 0.0086 Tw (date C by using a neural network.) Tj 147.84 0 TD 0 Tc 2.6311 Tw ( ) Tj -147.84 -11.52 TD ( ) Tj 0 -11.52 TD -0.0056 Tc -0.0033 Tw (Firstly, the historical data is prepared. At each time step, data are classified into one of two categories ) Tj 0 -11.4 TD -0.0097 Tc 0.0009 Tw (according to whether it was profitable to buy or sell stock A at price B on date C:) Tj 351.84 0 TD 0 Tc 2.6311 Tw ( ) Tj -351.84 -11.52 TD ( ) Tj 0 -11.52 TD -0.0133 Tc 0 Tw (1.) Tj 8.28 0 TD 0 Tc 6.9511 Tw ( ) Tj 9.72 0 TD -0.016 Tc 0.0071 Tw (Long: ) Tj 27.6 0 TD -0.0107 Tc 0.0019 Tw ( buy the stock on date C.) Tj 109.92 0 TD 0 Tc 2.6311 Tw ( ) Tj -155.52 -11.52 TD -0.0133 Tc 0 Tw (2.) Tj 8.28 0 TD 0 Tc 6.9511 Tw ( ) Tj 9.72 0 TD 0 Tc -0.0096 Tw (Short: sell the stock on date C.) Tj 138.36 0 TD 0 Tc 2.6311 Tw ( ) Tj -156.36 -11.52 TD ( ) Tj 0 -11.52 TD -0.0063 Tc -0.0026 Tw (Having fitted a model with this historical data, the model can be used to predict a profitable position at ) Tj 0 -11.4 TD -0.0019 Tc -0.007 Tw (time t+1 \(e.g., the next day or week\). At the end of each time step the model is update) Tj 380.88 0 TD -0.0155 Tc 0.0067 Tw (d to include the ) Tj -380.88 -11.52 TD -0.015 Tc 0.0061 Tw (new historical data.) Tj 84.96 0 TD 0 Tc 2.6311 Tw ( ) Tj -84.96 -11.52 TD ( ) Tj 0 -11.52 TD -0.0102 Tc 0.0013 Tw (By the time date C arrives, the trader should be in a profitable position \(either long or short\) given the ) Tj T* 0.0236 Tc -0.0325 Tw (current market value of stock A.) Tj 140.64 0 TD 0 Tc 2.6311 Tw ( ) Tj -140.64 -12.48 TD /F2 11.04 Tf 2.8109 Tw ( ) Tj 0 -12.48 TD /F3 11.04 Tf -0.0124 Tc 0.0633 Tw (Trading rules) Tj 69.84 0 TD 0 Tc 3.0509 Tw ( ) Tj -69.84 -12.72 TD /F2 11.04 Tf 2.8109 Tw ( ) Tj 0 -11.76 TD /F2 9.96 Tf 0.0046 Tc -0.0135 Tw (Trading rules can be determined from data with a categorical ) Tj 270.84 0 TD -0.0059 Tc -0.003 Tw (outcome, e.g., buy or sell, rise or fall. Such ) Tj -270.84 -11.52 TD 0.008 Tc -0.0168 Tw (rules take the form of a set of conditional statements and an action, e.g.,) Tj 318.72 0 TD 0 Tc 2.6311 Tw ( ) Tj -318.72 -11.4 TD 33.2311 Tw ( ) Tj 36 0 TD 0.0232 Tc -0.0321 Tw (IF CONDITION1 AND CONDITION 2 THEN ACTION) Tj 234.84 0 TD 0 Tc 2.6311 Tw ( ) Tj -270.84 -11.52 TD -0.0115 Tc 0.0027 Tw (and can be found by viewing a fitted decision tree model. Given an appropriate historica) Tj 388.2 0 TD -0.0082 Tc -0 Tw (l data set, this ) Tj -388.2 -11.52 TD -0.0105 Tc 0.0016 Tw (approach could be used to either validate a rule thought to exist, or generate new rules and ideas.) Tj 429.72 0 TD 0 Tc 2.6311 Tw ( ) Tj -429.72 -11.52 TD ( ) Tj 0 -11.52 TD -0.0077 Tc -0.0012 Tw (Suppose that a decision is fitted using historical data. Each internal node in the tree is a test on one of ) Tj T* -0.0082 Tc -0 Tw (the variables used to predict) Tj 124.32 0 TD -0.0043 Tc -0.0046 Tw ( the outcome in the historical data. If the variable, say X1, takes continuous ) Tj -124.32 -11.4 TD -0.0074 Tc -0.0015 Tw (values, this test is either:) Tj 108.84 0 TD 0 Tc 2.6311 Tw ( ) Tj -108.84 -11.52 TD 33.2311 Tw ( ) Tj 36 0 TD -0.0058 Tc -0.0031 Tw (\(X1 >= VALUE\) or \(X1 < VALUE\),) Tj 148.44 0 TD 0 Tc 2.6311 Tw ( ) Tj ET endstream endobj 15 0 obj << /Resources << /Font << /F4 16 0 R /F3 4 0 R /F2 5 0 R /F1 6 0 R /F0 8 0 R >> /ProcSet 10 0 R /XObject << /im2 17 0 R >> >> /Type /Page /Parent 1 0 R /Contents [18 0 R 19 0 R] /MediaBox [0 0 612 792] >> endobj 1 0 obj << /Kids [3 0 R 15 0 R] /Type /Pages /Count 2 >> endobj 20 0 obj << /Type /Catalog /Pages 1 0 R >> endobj 21 0 obj << /ModDate (D:20170713211420+02'00') /CreationDate (D:20170713211420+02'00') /Creator (pdftk 2.01 - www.pdftk.com) /Producer (itext-paulo-155 \(itextpdf.sf.net-lowagie.com\)) >> endobj xref 0 22 0000000000 65535 f 0000054007 00000 n 0000000000 65535 f 0000028483 00000 n 0000000015 00000 n 0000000128 00000 n 0000000489 00000 n 0000000236 00000 n 0000001824 00000 n 0000001561 00000 n 0000002901 00000 n 0000002939 00000 n 0000020335 00000 n 0000021056 00000 n 0000000000 65535 f 0000053781 00000 n 0000028697 00000 n 0000028813 00000 n 0000046209 00000 n 0000046930 00000 n 0000054073 00000 n 0000054125 00000 n trailer << /Info 21 0 R /ID [<53f9c8f5de85b603e277c6ab918bca8c><7ac8b010af0cac0fec936b8ce2f3775b>] /Root 20 0 R /Size 22 >> startxref 54321 %%EOF