2d Iteratioal Coferece o Electrical, Computer Egieerig ad Electroics (ICECEE 2015) Research o the Classificatio Algorithms for the Classical Poetry Artistic Coceptio based o Feature Clusterig Methodology Ji-feg LIANG 1, a 1 Guagi Techical College of Machiery ad electricty, Naig 530007, Chia a liagjifeg@126.cm Keywords: Classificatio Algorithm; Classical Poetry Artistic Coceptio; Fuzzy Clusterig. Abstract. From aciet to owadays, myriad literati ad poets had wrote may popular works, these works is the pearl of Chiese literature ad treasures of traditioal culture. Due to the legth limit of classical poetry, there ofte appear a lot of allusios. I this paper, we coduct theoretical aalysis ad umerical simulatio o the classificatio algorithms for the classical poetry artistic coceptio based o feature clusterig methodology. Artistic coceptio of Chiese classical poetry is a importat aesthetic category i the theory which is also a importat topic of Chiese traditioal aesthetics. It is ecessary to improve the classical tet classificatio model ad the applied to the classificatio of Chiese classical poetry. By a large umber of classical poetry ca be foud after readig ad classical Chiese poems are ofte ot directly vetilatio thought emotio. Our proposed method performs well i the eperimetal sectio. Itroductio Chia has the value reputatio ad otatio of poetry kigdom. From aciet to owadays, myriad literati ad poets had wrote may popular works, these works is the pearl of Chiese literature ad treasures of traditioal culture. Due to the legth limit of classical poetry, there ofte appear a lot of allusios, ligo i his works ad the cotet is ofte difficult to uderstad for ordiary readers so for a log time, the study of classical poetry is ofte some eperts ad scholars i this field has deep attaimets. These eperts ad scholars ofte kowledgeable will appear i the poetry of some specific cotet has the kee isight. For automatic classificatio system, its mai characteristic is based o statistical method ad based o machie learig method is applied to automatic tet classificatio. Amog them ad based o statistical method is carried out o the characteristics of the term i the documet statistics through the statistical results to aalyze the cotet of the documet. The method based o machie learig is to use the characteristics of the computig speed is fast to a large umber of statistical cotet through the computer to fid the coectio betwee them ad thus obtais the coclusio of some ordiary people is ot so easy to fid. So will be based o the statistical method combied with machie learig to deal with the tet for classificatio is a good way. Based o statistical classificatio method is, i essece, a quatitative reasoig uder ucertaity. It through the statistics ad aalysis o the large-scale corpus iformatio eeded to gai kowledge so that they ca o the whole documet of after data processig provides more objective data quality with high coverage ad good cosistecy. Of course it may igore the smaller as a proportio of the documet category. For algorithm based o coected it should be able to improve the parallelism of computatio ad the realizatio of the distributio of iformatio access advatage ad also has high fault tolerace for we have to lear the comple oliear mappig is a good choice. Therefore, to deal with the metioed challeges ad drawbacks, we coduct theoretical aalysis ad eperimetal implemetatio of the classificatio algorithms for the classical poetry artistic coceptio based o feature clusterig methodology. Through sort of classical poetry ito the problem of tet categorizatio we ca make use of some classical machie learig method, compared with geeral tet but Chiese classical poetry has a special characteristic especially compared with the moder tet differece is bigger. Therefore it is ecessary to improve the classical 2015. The authors - Published by Atlatis Press 423
tet classificatio model ad the applied to the classificatio of Chiese classical poetry. By a large umber of classical poetry ca be foud after readig, classical Chiese poems are ofte ot directly vetilatio thought emotio but by a word or words i the poem readers to depict the picture ad the sog poetry through these pictures coect. Whe readers after readig the whole works like the movies ad these pictures show readers dyamically so as to brig readers a kid of artistic coceptio, it is through this kid of artistic coceptio that readers feel wat to epress the cotet of the work. I the followig sectios we will discuss the algorithm i detail [1-4]. Our Desiged Algorithm ad Approach The Cocepts of Clusterig ad Classificatio. As we say, data aotatio is a additioal, epesive, ad error-proe preparatio process. Idividual data have to be carefully ispected i order to pipoit somewhat reliable class labels for the traiig patters. Istaces of the difficulties ivolved i the process are foud i areas such as bioiformatics, speech processig, or affective computig, where the eact class labels may ot eve be eplicitly observable. I the followig formula oe, we defie the data set for further processig. {(, ) d i i i, i, 1,2,..., } S= y R y Yi= m (1) Etry documet frequecy refers to the corpus i the terms of the umber of the documet. Documet frequecy is below a certai threshold of etry is low frequecy words ad will be removed from the origial feature space this etry will ot oly reduce the dimesio of feature space ad could improve the accuracy of classificatio. Mutual iformatio is widely used i statistical laguage model which is show below i the formula 2. MI t, c = log p t c / p t p c (2) ( ) ( ) 2 ( ( )) ( ( ) ( )) Multi-label learig is a kid of complicated decisio task ad the same object ca belog to multiple categories. At the same time such tasks i tet classificatio, image recogitio ad gee aalysis i areas such as widespread. Multi-label classificatio task ofte described by the high-dimesioal feature ad there are a lot of irrelevat ad redudat iformatio. Curretly has a large umber of sigle tag feature selectio algorithm is proposed to cope with the problem of dimesio disaster but for may tag attributes reductio ad feature selectio are rarely studied. Applyig rough set to multiple tags feature selectio ad the data for multi-label classificatio task which will be redefied the eighborhood rough set calculatio method of the lower approimatio ad depedece show i the formula 3 ad the table oe. U =,,..., ; A = a,a,...,a ;D = l,l,...,l (3) { } { } { } 1 2 1 2 1 2 Table.1The Eample of Sigle-Label Task for Clusterig a a 2 a 3 D 1-1 0 0 l 1 2 1 0 1 l 2 3 2 1 1 l 3 4 3 2-1 l 4 U 1 From the perspective of classificatio, patter recogitio is oe kid of the specific thigs to specific process. That is to use a certai umber of samples, classifier desig accordig to the similarity betwee them ad the use the desiged classifier to idetify the samples classificatio decisios. Classificatio process ca be either i the origial data space ad also ca to trasform the origial data the data image to the most ca reflect the ature of classificatio i the feature space. I cotrast, the latter makes it easier to decisio the desig of the machie ad it said through a more stable characteristics which will improve the decisio-makig performace of the machie ad 424
remove redudat or irrelevat iformatio ad more easy to fid the iheret relatioship betwee the object of study. The followig epressio shows the process. (, ) ( ) ( ) ( ) ( ) ( ) f C t = g C p t g C t + p t g C t (4) Aother key task of patter recogitio is for data aalysis ad processig. I geeral, the patter recogitio system described by the characteristics ad attributes cotaied i the object of the mathematical model of object of study i geeral depeds o the mutual relatios betwee characteristics. I cogitive sciece, the relatioship betwee objects is kow a lot of situatios ad the cause of sigificat differeces betwee the objects are ukow. The Priciple of Classical Poetry Artistic Coceptio. Of classical Chiese poetry i the high artistic achievemet, the roll degree we should to cherish today ad look at the aciets i poetry creatio eperiece ad aesthetic. Artistic coceptio of Chiese classical poetry is a importat aesthetic category i the theory which is also a importat topic of Chiese traditioal aesthetics. We from the creative practice of classical poetry ad cotact the aciet literary theory which ca be i the vast world to sum up the eperieces of aciet poets to create artistic coceptio of art ad eplore classical poetry show the art law of artistic coceptio. I the objective world, the object is specific ad limited, but the poet through the selectio of image ad a orderly combiatio of image creatio of artistic coceptio is a visioary blurred, is edless. Poet i choosig objective image build stereo space artistic coceptio of classical poetry, because of the limitatios of objective image ad artistic coceptio i the works of classical poetry ifiite etesibility, classical poetry artistic coceptio is boud to produce may do't eed to use a specific object caot use to fill the blak of specific image, thus formig the classical poetry artistic coceptio of the actual situatio of poetic ad artistic coceptio of poetic space. Our acestors of the traditioal thikig habits are mostly heavy ituitio, eperiece, to itegrity, comprehesive mode of thikig. I the whole, comprehesive thikig mode, which specializes i classical poetry creatio, is characterized by simple would be complicated laguage, dapper style ad rich implicatio betwee the height of the cojuctio. All works of artistic coceptio, should both do scee ad do the boudary begets like outside, false or true which give a perso leave a imagiatio that several layers of meaig are derived from reality to develop aesthetic imagiatio. Artistic coceptio is the lad bor like begets the outside, the actual product. Geerally speakig, the virtual eviromet is the sublimatio of reality, it reflects is habitat creatio itetio ad purpose, embodies the artistic coceptio of art taste ad aesthetic effect, restrictig is creatio ad descriptio, was the leader of the soul ad structure of artistic coceptio. But, small virtual eviromet ca be bor out of thi air, i the artistic coceptio creatio, everythig must also implemet to the reality of the figures. The Clusterig based Classificatio Algorithms for Coceptio. Most of eistig tet feature selectio methods are serial ad are iefficiet timely to be applied to Chiese massive tet data set. So, it is a hotspot of tet miig how to improve efficiecy of tet feature selectio by meas of parallel thikig. Combiig geetic algorithm with parallel collaborative evolutioary, a parallel collaborative evolutioary geetic algorithm based o rough sets was desiged ad used to select tet features. Each feature there is a certai tedecy, amely whe readig a poem poetry works, through to the idividual characteristics of artistic coceptio of eperiece ad readers will be easier to work with certai category associated to epress emotio. Is this effect, it is because if the characteristics of a work item oe aspect of the artistic coceptio, eviromet readers will aturally associated with this aspect. The type strig similarity calculatio method is show below. ( W W ) cos, 1 2 = w w 1i 2i i= 1 2 2 ( w1 i) ( w2i) i= 1 i= 1 Hierarchical clusterig method is also a kid of commo clusterig algorithm ad for the eed for clusterig of data accordig to the level of the steps. Accordig to the differet methods of (5) 425
hierarchical decompositio, hierarchical clusterig method ca be divided ito a bottom-up ad top-dow clusterig. For every feature of the artistic coceptio with the characteristics of a certai tedecy of feature represetatio method is improved, puts forward the mode with vector to represet the feature, vector of each compoet represets the feature weight i each category. After the aalysis from the perspective of the reader, the reader ca feel the sog poetry as a whole by the artistic coceptio, ad from the coceptio to covey the author wats to covey thoughts ad feeligs, based o this proposed to the characteristics of clusterig, o the oe had, those with similar artistic feature together, to achieve the purpose of sematic cotet retai more poetry, o the other had, the clusterig vector ca be dimesio reductio, improve the classificatio performace of the algorithm. Mea shift is calculated at the curret poit as the ceter ad the most populous area aroud it ad the move the ceter of the curret the most heavily populated areas after repeated iterative calculatio whe the ceter was o loger drift or drift rage withi the scope of the threshold value ad the you ca get the largest local desity of poits. Eperimet ad Simulatio Result I this sectio, we coduct umerical simulatio o the proposed methodology. The data is captured from the google scholar search egie. We use the stadard library to test the effectiveess of the proposed approach. The eperimetal result idicates that our method performs well. Coclusio ad Summary Fig. 1. The Eperimetal Result for the Proposed Algorithm I this paper, we coduct theoretical aalysis ad eperimetal implemetatio of the classificatio algorithms for the classical poetry artistic coceptio based o feature clusterig methodology. Through sort of classical poetry ito the problem of tet categorizatio we ca make use of some classical machie learig method, compared with geeral tet but Chiese classical poetry has a special characteristic especially compared with the moder tet differece is bigger. Based o statistical classificatio method is, i essece, a quatitative reasoig uder ucertaity. It through the statistics ad aalysis o the large-scale corpus iformatio eeded to gai kowledge so that they ca o the whole documet of after data processig provides more objective data quality with high coverage ad good cosistecy. The result proves the effectiveess of the proposed methodology ad i the future, we pla to coduct more research to polish the curret algorithm. 426
Refereces [1] He, P., Xie, G., Salamatia, K., & Mathy, L. (2014). Meta-algorithms for Software-Based Packet Classificatio. IEEE Iteratioal Coferece o Network Protocols (pp.308-319). IEEE. [2] Trappey A J C, Trappey C V, Hsu F C, et al. A fuzzy otological kowledge documet clusterig methodology[j]. Systems, Ma, ad Cyberetics, Part B: Cyberetics, IEEE Trasactios o, 2009, 39(3): 806-814. [3] Sog Q, Ni J, Wag G. A fast clusterig-based feature subset selectio algorithm for high-dimesioal data[j]. Kowledge ad Data Egieerig, IEEE Trasactios o, 2013, 25(1): 1-14. [4] Boţ R I, Heirich A, Waka G. Employig differet loss fuctios for the classificatio of images via supervised learig[j]. Cetral Europea Joural of Mathematics, 2014, 12(2):381-394. [5] Keylock C J, Sigh A, Veditti J G, et al. Robust classificatio for the joit velocity-itermittecy structure of turbulet flow over fied ad mobile bedforms[j]. Earth Surface Processes & Ladforms, 2014, 39(13):1717 1728. [6] Boutemedjet S, Bouguila N, Ziou D. A hybrid feature etractio selectio approach for high-dimesioal o-gaussia data clusterig[j]. Patter Aalysis ad Machie Itelligece, IEEE Trasactios o, 2009, 31(8): 1429-1443. [7] Evebly G, Vidal G. Algorithms for etaglemet reormalizatio: boudaries, impurities ad iterfaces[j]. Joural of Statistical Physics, 2013, 157(4-5):931-978. 427