Advaced Scece ad Techology Letters Vol.3 (SoftTech 06, pp.59-63 http://dx.do.org/0.457/astl.06.3.3 Research o SVM Predcto Model Based o Chaos Theory Sog Lagog, Wu Hux, Zhag Zezhog 3, College of Iformato Egeerg, North Cha Uversty of Water Resources ad Electrc Power, Zhegzhou Hea, 4500 Cha College of Iformato Egeerg, North Cha Uversty of Water Resources ad Electrc Power, Zhegzhou Hea, 4500 Cha 3 School of Water Coservacy, North Cha Uversty of Water Resources ad Electrc Power, Zhegzhou Hea, 4500 Cha Abstract. I order to ehace predcto precso of ole publc opo, t put forward a kd of ole publc opo predcto model (PSR-SVR wth the combato of chaos theory ad support vector regresso. Frst of all, the orgal data of ole publc opo were obtaed throughout topc segmetato, hotspot extracto, ad data aggregate. The, tme sequece of ole publc opo was recostructed throughout phase-space recostructo. Fally, the recostructed tme sequece of ole publc opo was put support vector regresso for modelg ad predcto, ad the t was compared wth other ole publc opo predcto model by expermet. The result shows that compared wth the cotrast model, PSR-SVR mproves the predcto precso ad relablty of ole publc opo, ad the predcto results have certa practcal value. Keywords: Ole publc opo; Support vector regresso; Phase space recostructo; Chaos theory Itroducto Ole publc opo s a mportat part of socal publc opo; compared wth the tradtoal ews meda, t's hghly teractve. The user s ot oly the formato recever, but also the formato source, whch makes the formato more tmely ad quckly spread o the Iteret. Negatve ole publc opo wll brg a greater threat to socal ad publc securty, so the aalyss ad modelg of the chages of ole publc opo, ad forecastg ts developg tred ca help the relevat departmets to formulate the correct to gude publc opo, ad has mportat practcal sgfcace to mata socal harmoy ad stablty [, ]. ISSN: 87-33 ASTL Copyrght 06 SERSC
Advaced Scece ad Techology Letters Vol.3 (SoftTech 06 Phase Space Recostructo ad Support Vector Regresso. Phase Space Recostructo Phase space recostructo theory s the bass of chaotc tme sequece predcto, ad the ma dea s: ay compoet evoluto of the system ad the teracto s determed by other compoets, ad ts related compoets formato s hdde the evoluto process, so through the aalyss of a certa compoet s tme sequece, the dyamc characterstcs of the orgal system ca be uderstood, wth the extracto ad restore of the orgal system []. For tme sequece of ole publc opo, x (t, t =,,..., N, by selectg a approprate embeddg dmeso m ad tme delayτ, tme sequece ca be recostructed, get a set of mult-dmesoal vector sequece of the formula (, so as to excavate the tme sequece hdde the ole publc opo, ad restore ole publc opo motve power system. X ( t x( t, x( t,, x[ t ( m ] ( Where, M=N-(m-τ. Ths paper uses the method of mutual formato ad G - P method to calculate ad determe the ole publc opo τ ad m.. Support Vector Regresso Support vector mache s a kd of mache learg algorthm based o statstcal learg theory, to fd the best compromse betwee the model complexty ad the ablty to lear, order to get the best geeralzato ablty [3]. The SVR regresso estmate fucto s f ( x w( x b ( Where, w s weght vector, b s the bas vector. makg the predctve expected rsk fucto mmum m J w C ( (3 costrat codto s: 60 Copyrght 06 SERSC
Advaced Scece ad Techology Letters Vol.3 (SoftTech 06 y w( x b w ( x b y, 0,,,, (4 Where,, s relaxg factor, C s pealty factor. Throughout troducg Lagrage multpler, the above optmzato problem s chages as typcal covex quadratc optmzato problem, amely, L w b w C (,,,,,,, ( ( y f ( x ( y f ( x ( (5 Where, ad represet Lagrage multpler. W (, ( ( j j ( ( x,, j ( xj ( y ( (6 costrat codto s: w ( x, j ( 0 (7 0, C Upo solvg the practcal questos, t oly eeds to make use of support vector mache to get the soluto, so the regresso estmato fucto s Copyrght 06 SERSC 6
Advaced Scece ad Techology Letters Vol.3 (SoftTech 06 ( ( ( (, ( f x x x b (8 Kerel fucto k( x, x s adapted to replace ( ( x, ( x, whch ca avod curse of dmesoalty, so ( ( (, f x k x x b (9 Ths paper selects radal bass fucto kerel as kerel fucto of SVR, fally regresso fucto of SVR: f ( x exp N x xj b (0 Where, s the wdth of radal bass fucto kerel [3]. 3 Cocluso The ole publc opo s affected by varous factors, ad s characterzed by tmevaryg, chaos. It s a kd of complex chages system, ad the tradtoal predcto algorthm s dffcult to establsh accurate predcto model. Accordg to the chaotc characterstcs of ole publc opo chages, by usg chaos theory ad SVR, a model of ole publc opo predcto based o PSR SVR was bult. Results show that compared wth cotrast model, PSR - SVR mproves the predcto precso of ole publc opo, predcto results are more stable, ad t more accurately descrbes the complex chage tred of the ole publc opo. The predcto results are helpful to correctly uderstad the developmet of ole publc opo, thus helpg to scetfcally gude ad maage varous ole publc opo trasmsso platforms, to promote the work of buldg a harmoous socety. Ackowledgmets. Ths work was supported by the Natoal Natural Scece Foudato of Cha (5309098. 6 Copyrght 06 SERSC
Advaced Scece ad Techology Letters Vol.3 (SoftTech 06 Refereces. Hu, J., Gao Z., Pa, W.: Multagle Socal Network Recommedato Algorthms ad Smlarty Network Evaluato [J]. Joural of Appled Mathematcs, 03 (03. Lv, Z., Y, T., Ha, Y., Che, Y., Che, G.: WebVR web vrtual realty ege based o PP etwork. Joural of Networks. 6, o. 7 (0: 990-998 3. Yag, J., Che, B., Zhou, J., Lv, Z.: A portable bomedcal devce for respratory motorg wth a stable power source. Sesors. (05 4. Zhao, D.: FusoFS: Toward supportg data-tesve scetfc applcatos o extremescale hgh-performace computg systems. Bg Data (Bg Data, 04 IEEE Iteratoal Coferece o. IEEE, (04 5. Dag, S., Ju, J., Matthews, D., Feg, X., Zuo. C.: Effcet solar power heatg system based o letcular codesato. Iformato Scece, Electrocs ad Electrcal Egeerg (ISEEE, 04 Iteratoal Coferece o. 6-8 Aprl (04 Copyrght 06 SERSC 63