Yan et al. / J Zhejiang Univ-Sci C (Comput & Electron) in press 1. Improving Naive Bayes classifier by dividing its decision regions *

Size: px
Start display at page:

Download "Yan et al. / J Zhejiang Univ-Sci C (Comput & Electron) in press 1. Improving Naive Bayes classifier by dividing its decision regions *"

Transcription

1 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press 1 Journal of Zhejang Unversty-SCIENCE C (Computers & Electroncs) ISSN (Prnt); ISSN X (Onlne) E-mal: jzus@zju.edu.cn Improvng Nave Bayes classfer by dvdng ts decson regons * Zh-yong YAN, Cong-fu XU, Yun-he PAN (Insttute of Artfcal Intellgence, Zhejang Unversty, Hangzhou 3127, Chna) E-mal: xucongfu@zju.edu.cn Receved Dec. 2, 21; Revson accepted Apr. 8, 211; Crosschecked Abstract: Classfers can be regarded as dvdng the data space nto decson regons separated by decson boundares. In ths paper we analyze decson tree algorthms and the Tree algorthm from ths perspectve. A decson tree can be regarded as a classfer tree. Each classfer on a non-root node can be regarded as traned n decson regons of the classfer on the parent node. The Tree algorthm generates a classfer tree wth the C4.5 algorthm and the Nave Bayes classfer as the root and leaf classfers respectvely. The Tree algorthm can be regarded as tranng Nave Bayes classfers n decson regons of the C4.5 algorthm. We propose the second dvson (SD) and three soft second dvson (SD-soft) algorthms to tran classfers n decson regons of the Nave Bayes classfer. These four new algorthms all generate two-level classfer trees wth the Nave Bayes classfer as root classfers. The SD and three SD-soft algorthms can make good use of the nformaton contaned n nstances near decson boundares, and the nformaton may be gnored by the Nave Bayes classfer. Expermental results on 3 UCI data sets ndcate that when usng the C4.5 algorthm and support vector machne as leaf classfers, the SD algorthm can obtan better generalzaton abltes than the Tree and the AODE algorthms. Expermental results also ndcate that when selectng approprate argument values three SD-soft algorthms can obtan better generalzaton abltes than the SD algorthm. Key words: Nave Bayes classfer, Decson regon, Tree, C4.5 algorthm, SVM do:1.1631/jzus.c1437 Document code: A CLC number: 1 Introducton The Nave Bayes classfer (Domngos and Pazzan, 1997) s an example of global learnng (Huang et al., 28), whch obtans a dstrbuton estmaton of the whole data set. The Nave Bayes classfer assumes that attrbutes of nstances are ndependent gven the class (Domngos and Pazzan, 1997). Although ths assumpton s very nave, the Nave Bayes classfer has good generalzaton ablty, and s one of top 1 algorthms n data mnng voted by IEEE ICDM 26 (Wu et al., 28). There are many studes on mprovng the generalzaton ablty of the Nave Bayes classfer, amongst whch the Tree algorthm (Kohav, 1996) Correspondng author * Ths paper s supported by the Natonal Natural Scence Foundaton of Chna under Grant No , and partally by the Natonal Basc Research Program of Chna under Grant No. 21CB Zhejang Unversty and Sprnger-Verlag Berln Hedelberg 211 s typcal. The Tree algorthm trans Nave Bayes classfers on the leaf nodes of a decson tree. Some researchers regard classfcaton as dvdng data space X nto some decson regons separated by decson boundares (Bshop, 26), although most researchers regard classfcaton as fndng a mappng from data space X to label set Y (Mtchell, 1997). We call the former perspectve as the dvdng perspectve. In ths paper, we wll analyze decson tree algorthms and the Tree algorthm from the dvdng perspectve. A decson tree can be regarded as a classfer tree composed of two types of classfers. An Tree can be regarded as a two-level classfer tree wth a decson tree classfer as the root node and several Nave Bayes classfers as leaf nodes. Then the Tree algorthm can be regarded as tranng Nave Bayes classfers n decson regons of the C4.5 algorthm (Qunlan, 1993). The Tree algorthm can utlze advantages of both the C4.5 algorthm and the Nave Bayes classfer, and t outper-

2 2 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press forms these two classfers (Kohav, 1996). Beng contrary to the Tree algorthm whch trans Nave Bayes classfers n decson regons of the C4.5 algorthm, we propose the second dvson (SD) algorthm to tran classfers n decson regons of the Nave Bayes classfer. The SD algorthm wll generate a two-level classfer tree, wth the Nave Bayes classfer as the root node and other classfers as leaf nodes. We also propose three soft versons of the SD algorthm (SD-soft) to deal wth overlapped regons generated by the Nave Bayes classfer. In ths paper leaf classfers used are the Nave Bayes classfer, the C4.5 algorthm and Support Vector Machne (SVM) (Vapnk, 1995). We perform experments on 3 UCI data sets (Murphy and Aha, 1998) to compare the SD algorthm wth the Tree and the AODE algorthms (Webb et al., 25). We also compare three SD-soft algorthms wth the SD algorthm. We apply the SD algorthm to the AODE algorthm, the C4.5 algorthm and SVM. Fnally we adopt the global / local learnng theory of Huang et al (28) to dscuss why the SD algorthm works. 2 Analyss of decson tree and the Tree algorthms from the dvdng perspectve The dvdng perspectve regards classfers as dvdng data space X nto some decson regons (Bshop, 26). The defnton of decson regon s as follows. Defnton 1 (Decson Regon) If regon R n data space X satsfes the followng two condtons for classfer C, then t wll be called the decson regon of label Y under classfer C, denoted as DR(C, Y ). x R, C( x ) = Y ; (1) x R, C( x ) Y. (2) For decson regons, the followng formulas are true. DR( CY, ), Y Y; (3) DR( CY, ) DRCY (, j) =, Y Yj; (4) DR( CY, ) = X. (5) Y Y From Eq. (3) (5), t s clear that decson regons of classfer C consttute a partton of data space X. From the dvdng perspectve, a classfer can be regarded as a dvder, whose work s to obtan a partton of data space X. 2.1 Analyss of decson tree algorthms The decson tree s a knowledge representaton method. Non-leaf nodes of a decson tree are attrbutes, and leaf nodes of a decson tree are labels. On each non-leaf node, branches are generated accordng to value of the attrbute on ths node. An example of the decson tree s shown n Fg. 1. R 1 N 1 X N R 2 R 3 R 4 R 5 R 6 N2 Y Y 1 Y 2 Y 3 Fg. 1 Structure of a decson tree and regons correspondng to ts nodes Each node of a decson tree corresponds to a regon of data space X. For example, the root node N of the decson tree n Fg. 1 corresponds to data space X. A leaf node of a decson tree corresponds to a part of a decson regon. Decson tree algorthms adopt a majorty votng method to determne the label for the regon correspondng to a leaf node. On each non-leaf node, decson tree algorthms dvde the regon to whch the node corresponds nto several regons to whch the node s chldren nodes correspond. The relatonshp among regons of parents and chldren nodes of the decson tree n Fg. 1 s as follows. X = R1 R2, R1 R2 =, R1, R2 ; (6) R1 = R3 R4, R3 R4 =, R3, R4 ; (7) R = R R, R R =, R, R. (8) It s clear that each non-leaf node corresponds to a dvder. From the dvdng perspectve, a classfer

3 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press 3 can be regarded as a dvder. If regons generated by a dvder are assocated wth labels, then regons can be regarded as decson regons. After ths process, a dvder can be regarded as a classfer. For example, f node N 1 and N 2 are assocated wth label Y 1 and Y 2 respectvely, then the root node N can be regarded as correspondng to a classfer, whose output s Y 1 or Y 2. Each non-leaf node corresponds to a pece-wse classfer (PWC) n Eq. (9). Obvously, PWC s a very smple classfer. Y, x[ ] < v1; Y1, x[ ] [ v1, v2); PWC( x) =... Yk 1, x[ ] vk 1. (9) The regon correspondng to a node can be regarded as a decson regon of the classfer correspondng to the parent node. For a non-leaf node, decson tree algorthms tran a PWC n the decson regon of the classfer correspondng to ts parent node. In Fg. 1, PWC correspondng to the root node N dvdes data space X nto R 1 and R 2, whch correspond to node N 1 and N 2 respectvely. For node N 1, PWC 1 s traned n decson regon R 1 of PWC. Smlarly, for node N 2, PWC 2 s traned n decson regon R 2 of PWC. Each leaf node corresponds to a majorty votng classfer (MVC) n Eq. (1). Lke PWC, MVC s also a very smple classfer. MVC( x) = argmax {( x, y) y = Y }. (1) Y Decson tree algorthms tran an MVC n the regon correspondng to a leaf node. The label predcted by a leaf node s the one predcted by the MVC traned n the regon correspondng to the leaf node. Each non-leaf node corresponds to a PWC, and each leaf node corresponds to an MVC, thus a decson tree can be regarded as a classfer tree. The classfer tree correspondng to the decson tree n Fg. 1 s shown n Fg. 2. Although both PWC and MVC are very smple, we can obtan very good generalzaton ablty by organzng these two classfers as a tree. For example, the C4.5 algorthm s one of top 1 algorthms n data mnng voted by ICDM 26 (Wu et al., 28). Fg. 2 The classfer tree correspondng to the decson tree n Fg Analyss of the Tree algorthm The Tree algorthm also generates a decson tree, whose leaf nodes are Nave Bayes classfers nstead of labels. An Tree s also a classfer tree, whose non-leaf and leaf nodes are PWCs and Nave Bayes classfers respectvely. The structure of an Tree s shown n Fg. 3. X PWC R 1 R 2 PWC 1 PWC 2 R 3 R 4 R 5 R Fg. 3 An Tree as a classfer tree In Fg. 3, f node, 1, 2 and 3 are assocated wth label Y, Y 1, Y 2 and Y 3, then regon R 3, R 4, R 5 and R 6 can be regarded as decson regons of PWC 1 and PWC 2. Then the sub-tree composed of PWC, PWC 1 and PWC 2 can be regarded as a decson tree generated by the C4.5 algorthm, whose objectve s to generate decson regons for tranng Nave Bayes classfers. So an Tree can also be regarded as a classfer tree composed of a C4.5 classfer and several Nave Bayes classfers. These Nave Bayes classfers are traned n decson regons of the C4.5 classfer. The two-level classfer tree correspondng to the classfer tree n Fg. 3 s shown n Fg. 4.

4 4 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press Fg. 4 An Tree s a classfer tree composed of a C4.5 and several s 3 Tranng classfers n decson regons of the Nave Bayes classfer Tranng Nave Bayes classfers n decson regons of the C4.5 algorthm can mprove the generalzaton abltes of both the Nave Bayes classfer and the C4.5 algorthm (Kohav, 1996). In ths secton, we wll study the method of tranng classfers n decson regons of the Nave Bayes classfer. 3.1 The SD algorthm The smplest method s to tran classfers n decson regons of the Nave Bayes classfer, whch s shown n Fg. 5. Fg. 5 Tran classfers n decson regons of the Nave Bayes classfer There are two questons when tranng classfers n decson regons of the Nave Bayes classfer: (1) When should we tran classfer C n a decson regon of the Nave Bayes classfer? (2) When classfer C s traned n a decson regon of the Nave Bayes classfer, there are two classfers n ths decson regon. Then whch classfer should be selected? For queston 1, t s clear that when classfer C can mprove the generalzaton ablty of the Nave Bayes classfer n the decson regon classfer C should be traned. However, t s very dffcult to determne whether classfer C can do ths. So we only exclude one mpossble case that when the tranng accuracy of the Nave Bayes classfer n the decson regon s 1%. When ths case s true, all nstances n the decson regon have the same class. If classfer C s traned n the decson regon, only a classfcaton model wth Y as ts unque output can be obtaned. Ths classfcaton model cannot mprove the generalzaton ablty of the Nave Bayes classfer n ths decson regon. For queston 2, between the Nave Bayes classfer and classfer C, the one wth better generalzaton ablty should be selected, but t s very dffcult to determne n the tranng phase. There are at least three methods for selectng a better classfer. The frst method s to dvde tranng data set D nto tranng sub-set and valdaton sub-set and then choose the classfer wth better test accuracy on the valdaton sub-set. The second method adopts cross valdaton to select the better classfer. The thrd method s to select the classfer wth better tranng accuracy. Amongst above three methods, the former two may obtan more accurate selecton, but they smultaneously need more tme. We adopt the last method n ths paper. Tranng the classfer can be regarded as dvdng data space X once. Then tranng classfer C n decson regons of the Nave Bayes classfer can be regarded as the second dvson to data space X. The procedure of the second dvson (SD) algorthm s shown n Fg. 6. Lke the Tree algorthm, the SD algorthm also generates a two-level classfer tree. In ths classfer tree, the Nave Bayes classfer s the root classfer, and classfer C s the leaf classfer. 3.2 The SD-soft algorthms The SD algorthm trans classfers n decson regons of the Nave Bayes classfer. Decson regons consttute a partton of data space X. Ths dvdng s hard dvdng. But the Nave Bayes classfer can generate overlapped regons n data space X, whch means nstance (x, y) may belong to several decson regons wth condton probabltes. In ths sub-secton we propose a soft verson of the SD al-

5 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press 5 gorthm to deal wth overlapped regons generated by the Nave Bayes classfer. Tranng phase: Input: tranng data set D, the Nave Bayes classfer, classfer C Output: a two-level classfer tree named SD tree (composed of, C, ndcator vector s_used) 1. Tran the Nave Bayes classfer on tranng data set D. 2. Obtan data sets {D } n decson regons of usng the followng formula. D = {( x, y) ( x ) = Y}. 3. For each D, compute tranng accuracy of. If t s 1%, s_used[] = false. If t s not 1%, tran C on D to obtan C. If tranng accuracy of C on D s greater than that of, then s_used[] = true, otherwse s_used[] = false. Test phase: Input: an SD tree, test nstance x Output: label y 1. Obtan predcton y of for x. 2. If s_used[y ]==false, y=y. Otherwse, y=cy (x). Fg. 6 Procedure of the SD algorthm There are many methods for generatng regons wth overlaps. In ths sub-secton, we adopt the method smlar to that whch the dvde-and-conquer algorthm (Frosynots et al., 23) uses to generate overlapped regons and combne classfers. The dvde-and-conquer algorthm frst uses a clusterng algorthm to obtan data sub-sets {D }, and then trans several mult-layered perceptron classfers (Pal and Mtra, 1992) on these data sub-sets. The clusterng algorthm used s fuzzy C-means algorthm (Bezdek, 1981) or greedy EM algorthm (Vlasss and Lkas, 22). In the test phase, predctons of these classfers are combned to make the fnal predcton. The membershp degree of nstance (x, y) belongng to cluster s denoted as u(cluster, x). Data sub-set D s obtaned by Eq. (11). D = {( x, y) u( cluster, x ) > q}. (11) The dvde-and-conquer algorthm then uses Eq. (12) to obtan the probablty dstrbuton p(y x). py ( x) = ucluster (, x) IC ( ( x ) == Y). (12) j j j= The ndcator functon I(z) s defned n Eq. (13). 1, f z s true, I( z) = (13), otherwse. We adopt the above sub-set generatng method to obtan regons wth overlaps. The output of the Nave Bayes classfer s a probablty dstrbuton vector [p (Y x),, p (Y k-1 x)]. The data sub-set D s obtaned by Eq. (14). D = {( x, y) p ( Y x ) > q}. (14) From Eq. (14), t s clear that nstance (x, y) may be contaned n several data sub-sets. Snce classfer C may be a classfer outputs the probablty dstrbuton, we mprove the combnng method of the dvde-and-conquer algorthm. Our combnng method s n Eq. (15). py ( x) = p ( Y x) p ( Y x ). (15) j Cj j= Takng ndcator vector s_used nto consderaton, the output of p(y x) s calculated by Eq. (16). py ( x) = p ( Y x) p ( Y x) I( jt, ) j Cj j= + p ( Y x) I(, F). (16) For smplcty, we use I(j, T) and I(j, F) to represent I(s_used[j]==true) and I(s_used[j]==false) respectvely. Then we have Eq. (17). I( j, T) + I(, F) = 1. (17) Eq. (18) s true for every classfer that outputs the probablty dstrbuton. pc( Y x ) = 1. (18) = Then we have the followng dervaton.

6 6 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press k 1 k 1 k 1 py ( x) = p ( Y x) p ( Y x) I( jt, ) j Cj = = j= + p ( Y x) I(, F) = k 1 k 1 = p ( Y x) I( j, T) p ( Y x) j C j j= = + p ( Y x) I(, F) = j j= = = p Y = = p ( Y x) I( j, T) + p ( Y x) I(, F) = p ( Y x) ( I( j, T) + I(, F)) = = 1 ( x) The fnal output s determned by Eq. (19). y = arg max p( Y x ). (19) Y Y Fnally, the soft verson of the SD algorthm (SD-soft) s shown n Fg Tme complexty of the SD algorthm We frst assume the tranng data has m attrbutes, n nstances and k classes. Then we dscuss the tme complexty of the SD algorthm. It s clear that the SD algorthm ncreases the computatonal cost of the Nave Bayes classfer.the tranng and test tme complextes of the Nave Bayes classfer are O(mn) and O(km) respectvely (Webb et al., 25). In the tranng phase, after tranng the Nave Bayes classfer, the SD algorthm frst dvdes the tranng data set nto data sub-sets, and the tme complexty s O(kmn). Then the SD algorthm trans classfer C at most k tmes, and the tme complexty s O(kC tran (n/k)), n whch O(C tran (n/k)) s the tme complexty of tranng classfer C on data set of sze n/k. Thus the tranng tme complexty of the SD algorthm s O(mn)+O(kmn)+O(kC tran (n/k)). In the test phase, the SD algorthm frst uses to obtan the predcton y, and the tme complexty s O(km). Then the SD algorthm may use classfer C y to make the fnal predcton, and the tme complexty s O(C test (n/k)), n whch O(C test (n/k)) s the tme complexty of usng classfer C to make the predcton. Thus the test tme complexty of the SD algorthm s O(km)+O(C test (n/k)). Tranng phase: Input: tranng data set D, the Nave Bayes classfer, classfer C, probablty threshold q. Output: a two-level classfer tree named SD tree (composed of, C, ndcator vector s_used) 1. Tran the Nave Bayes classfer on tranng data set D. 2. Obtan data sets {D s } and data set {D } n decson regons of usng followng formulas. Ds = {( x, y) p ( Y x ) > q}. D = {( x, y) ( x ) = Y}. 3. For each D, compute tranng accuracy of. If t s 1%, s_used[] = false. If t s not 1%, tran C on D s to obtan C. If tranng accuracy of C on D s greater than that of, then s_used[] = true, otherwse s_used[] = false. Test phase: Input: a SD tree, test nstance x Output: label y 1. Obtan predcton y of for x. 2. If s_used[y ]==false, y=y. Otherwse obtan probablty dstrbuton of for x, and then use followng two equatons to obtan the fnal predcton y. py ( x) = p ( Y x) p ( Y x) I( jt, ) j Cj j= + p ( Y x) I(, F). y = arg max p( Y x ). Y Y Fg. 7 Procedure of SD-soft algorthm In ths paper, we use the Nave Bayes classfer, the C4.5 algorthm and SVM as leaf classfers, and SVM adopts sequental mnmal optmzaton (SMO) tranng algorthm (Platt, 1999). These three classfers are eager classfers, whose test tme complexty s very small. Thus we only compare the tranng tme complexty of the SD algorthm wth the Nave Bayes classfers, the Tree and the AODE algorthms. Tranng tme complextes are lsted n Table 1. From Table 1, t s clear that tranng tme complextes of SD(C=) and SD(C=C4.5) are smaller than that of the Tree algorthm, and the tranng tme complexty of SD(C=SVM) may be comparable wth that of the Tree algorthm. The tranng tme

7 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press 7 complexty of SD(C=C4.5) may be comparable wth that of the AODE algorthm. The tranng tme complexty of SD(C=SVM) s larger than that of the AODE algorthm. The computatonal cost of the SD-soft algorthm s larger than that of the SD algorthm because data set D s usually has more nstances than data set D. Table 1 Tranng tme complextes of eght classfers Classfer Tme Complexty and Lterature O(mn) (Webb et al., 25) Tree O(n 2 km 2 /v) (Zheng and Webb, 25) AODE O(m 2 n) (Webb et al., 25) C4.5 O(mnlogn) (Frank and Wtten, 1998) SMO O(n d ), d [1, 2.2] (Platt, 1999) SD(C=) O(mn)+O(kmn)+ O(mn) SD(C=C4.5) O(mn)+O(kmn)+O(mnlog(n/k)) SD(C=SMO) O(mn)+O(kmn)+O(n d /k d-1 ) k s the number of classes. m s the number of attrbutes. v s the average number of values for an attrbute. n s the number of tranng nstances. 4 Expermental results In ths secton, we perform experments to use the SD and the SD-soft algorthms to tran the Nave Bayes classfer, the C4.5 algorthm and SVM n decson regons of the Nave Bayes classfer. We frst compare the SD-algorthm wth the Tree and the AODE algorthms, and then we compare the SD-soft algorthm wth the SD algorthm. The expermental envronment s WEKA (Wtten and Frank, 25). The Nave Bayes classfer, the Tree algorthm and the C4.5 algorthm adopt NaveBayes, Tree and J48 methods of WEKA respectvely, wth default settngs. J48 method s an mplementaton of the C4.5 release 8 (Qunlan, 1996). The AODE algorthm adopts AODE method of WEKA, and adopts the supervsed dscreton method to dscretze contnual attrbutes. SVM adopts SMO method of WEKA, and adopts a lnear kernel functon. The expermental data sets are 3 UCI data sets. We remove nstances wth unknown values from these data sets. After the above preprocessng, descrptons of these data sets are lsted n Table 2. Table 2 Descrptons of 3 UCI data sets after preprocessng Dataset No. of No. of No. of attrbutes classes nstances anneal balance-scale breast-cancer brdges car credt-g dermatology dabetes ecol haberman heart-c heart-statlog hepatts onosphere rs lver-dsorders lung-cancer molecular mushroom postoperatve segment solar sonar soybean sponge tc-tac-toe vehcle vote vowel zoo We adopt 1 tmes 1-fold cross valdaton to obtan test accuraces and adopt two-taled t-test wth 95% confdence to compare dfferent classfers to obtan wn-loss-te (w-l-t). Due to space lmtaton, only t-test results are lsted. 4.1 The SD algorthm We frst use the Nave Bayes classfer as the leaf classfer. The SD algorthm can be appled twce and three tmes, and they are denoted as 2SD and 3SD respectvely. We tran the Nave Bayes () classfer, SD, 2SD and 3SD, and then compare these classfers by t-test. The t-test results are lsted n Table 3.

8 8 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press Table 3 The t-test results (w-l-t) of applyng the SD algorthm once, twce and three tmes SD vs. 2SD vs. 3SD vs. 2SD vs. SD 3SD vs. 2SD From Table 3, SD, 2SD and 3SD all can mprove the generalzaton ablty of the Nave Bayes classfer. Compared wth the Nave Bayes classfer, 2SD obtans the best t-test result. Thus t s necessary to apply the SD algorthm twce. Then we compare 2SD wth the Tree and the AODE algorthms. Table 4 lsts the t-test results. Table 4 The t-test results (w-l-t) of comparng 2SD wth Tree and AODE 2SD vs. Tree 2SD vs. AODE From Table 4, 2SD can obtan better generalzaton ablty than the Tree algorthm, but the generalzaton ablty of 2SD s worse than that of the AODE algorthm. Another advantage of 2SD over the Tree algorthm s that the tranng tme of 2SD s much shorter than that of the Tree algorthm, because the SD algorthm does not adopt cross valdaton n the tranng phase. We then use the C4.5 algorthm and SVM as leaf classfers of the SD algorthm. Test accuraces obtaned by the SD algorthm are compared wth those obtaned by the Nave Bayes classfer, leaf classfers, the Tree and the AODE algorthms. The t-test results are lsted n Table 5. Table 5 The t-test results (w-l-t) of comparng the SD algorthm wth other four classfers SD C Tree AODE C=C C=SVM From table 5, t s clear when usng C4.5 algorthm and SVM as leaf classfers the SD algorthm can obtan better generalzaton abltes than the Nave Bayes classfer, leaf classfers, the Tree and the AODE algorthms. 4.2 The SD-soft algorthm In ths sub-secton, we compare the SD-soft algorthm wth the SD algorthm. The selecton of q value s crucal for the SD-soft algorthm. In our experment, the SD-soft algorthm s traned wth q values n Eq. (2). q= p/ m, p= {.,.2,.4,..., 2.}. (2) We tran the SD-soft algorthm wth 11 q values and the SD algorthm on 3 UCI data sets. The leaf classfers used are the Nave Bayes classfer, the C4.5 algorthm and SVM. We frst compare average test accuraces of the SD algorthm wth those of the SD-soft algorthm wth dfferent q values. Average test accuraces are shown n Fg. 8. The best q value we select for a data set s the one that makes the SD-soft algorthm obtan the best average test accuracy on ths data set. Then the SD-soft algorthm wth the best q value s compared wth the SD algorthm. The t-test results are lsted n Table 6. accuracy SD(C=) SD-soft (C=) SD(C=C4.5) SD-soft(C=C4.5) SD(C=SVM) SD-soft (C=SVM) p Fg. 8 Test accuraces of the SD and the SD-soft algorthms wth dfferent p values From Fg. 8, when usng the Nave Bayes classfer or SVM as the leaf classfer, the SD-soft algorthm obtans worse test accuraces than the SD algorthm. However when usng the C4.5 algorthm as the leaf classfer and when the p value s.6,.8 or 1., the SD-soft algorthm obtans better test accuraces. From Table 6, when the best average test accuracy s selected for each data set, the SD-soft algorthm can obtan better generalzaton ablty than the SD algorthm. Thus t s possble for the SD-soft al-

9 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press 9 gorthm to obtan better generalzaton ablty than the SD algorthm f approprate q value s selected. Table 6 The t-test results (w-l-t) of comparng the SD-soft algorthm wth the SD algorthm SD-soft C= C=C4.5 C=SVM t-test Dscussons In ths secton, we frst propose other two SD-soft algorthms. Then we apply the SD algorthm to the AODE algorthm, the C4.5 algorthm and SVM. Fnally we dscuss why the SD algorthm works. 5.1 Other two SD-soft algorthms The SD-soft algorthm s proposed to deal wth the overlapped regons generated by the Nave Bayes classfer. The SD-soft algorthm adopts the method whch the dvde-and-conquer algorthm uses to deal wth the overlapped regons. In ths sub-secton, we nvestgate two other methods of dealng wth the overlapped regons. We frst use Eq. (21) to obtan the data set {D s } of the SD-soft algorthm, and the resultng algorthm s named SD-soft.v2. When q.5, data set D s s the same as data set D. Thus when q.5, the SD-soft.v2 algorthm uses the predcton method of the SD algorthm to make the predcton. D = D {( x, y) ( x) Y, p ( Y x ) > q}. (21) s The SD-soft and the SD-soft.v2 algorthms are both based on the condtonal probablty p (Y x). Next we propose another SD-soft algorthm based on the predcton rank of label Y. The functon r(, x, Y ) (Eq. (22)) calculates the predcton rank of label Y. r (, x, Y) = { Y p ( Y x) > p ( Y x )} + 1. (22) j j We use Eq. (23) to obtan the data set {D s } of the SD-soft algorthm, and the resultng algorthm s named SD-soft.v3. When k=1, data set D s s the same as data set D. Thus when k=1, the SD-soft.v3 algorthm uses the predcton method of the SD algorthm to make the predcton. D = {( x, y) r(, x, Y) k, p ( Y x ) > }. (23) s We compare above two new SD-soft algorthms wth the SD algorthm. The SD-soft.v2 algorthm uses the q value n Eq. (2). The k values of the SD-soft.v3 algorthm are 1, 2 and 3. For these two new algorthms, we select the best average test accuracy amongst dfferent argument values for each data set, and compare t wth the test accuracy of the SD algorthm. The t-test results are lsted n Table 7. Table 7 The t-test results (w-l-t) of comparng two SD-soft algorthms wth the SD algorthm SD-soft C= C=C4.5 C=SVM v v From Table 7, these two new algorthms both can obtan better generalzaton abltes than the SD algorthm. Comparng Table 7 wth Table 6, we fnd that they cannot obtan better generalzaton abltes than the SD-soft algorthm, except that when the SD-soft.v3 algorthm adopts SVM as the leaf classfer. Lke the SD-soft algorthm, argument selecton s crucal for the above two new algorthms. 5.2 Applyng the SD algorthm to the AODE algorthm, the C4.5 algorthm and SVM We frst adopt the SD algorthm to tran the C4.5 algorthm and SVM n decson regons of the AODE algorthm. The t-test results are lsted n Table 8. Table 8 The t-test results (w-l-t) of applyng the SD algorthm to the AODE algorthm SD(AODE+C) vs. AODE vs. C C=C C=SVM From Table 8, tranng C4.5/SVM n decson regons of the AODE algorthm by the SD algorthm can mprove generalzaton abltes of both the AODE algorthm and leaf classfers. We then adopt the SD algorthm to tran the C4.5 algorthm, the Nave Bayes classfer and SVM n decson regons of the C4.5 algorthm and SVM. The t-test results are lsted n Table 9 and 1 respectvely.

10 1 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press Table 9 The t-test results (w-l-t) of applyng the SD algorthm to the C4.5 algorthm SD(C4.5+C) vs. C4.5 vs. C C=C C= C=SVM Table 1 The t-test results (w-l-t) of applyng the SD algorthm to SVM SD(SVM+C) vs. SVM vs. C C=SVM C=C C= From Table 9 and 1, the SD algorthm s applcable for nether the C4.5 algorthm nor SVM. SD(C4.5+C4.5) mproves the generalzaton ablty of C4.5 algorthm n only sx data sets. SD(C4.5+) and SD(C4.5+SVM) both mprove the generalzaton ablty of the C4.5 algorthm, but they reduce the generalzaton abltes of the Nave Bayes classfer and SVM. SD(SVM+SVM) mproves the generalzaton ablty of SVM n eght data sets, but reduces t n fve data sets. SD(SVM+C4.5) and SD(SVM+) cannot mprove the generalzaton ablty of SVM, although they mprove the generalzaton abltes of the C4.5 algorthm and the Nave Bayes classfer n over half the data sets. Thus the SD algorthm s applcable for the AODE algorthm, but t s applcable for nether the C4.5 algorthm nor SVM. 5.3 Why does the SD algorthm work? From expermental results of secton 4 and sub-secton 5.2, the SD algorthm s applcable for both the Nave Bayes classfer and the AODE algorthm, but t s not applcable for ether the C4.5 algorthm or SVM. In ths sub-secton we attempt to nvestgate the reason. Huang et al (28) dvdes classfers nto the global learnng and the local learnng. The global learnng makes descrptons of data, whereas the local learnng captures the local useful nformaton from data. The global learnng ams to estmate a dstrbuton of data, whereas the local learnng ams to obtan decson boundares for classfcaton. The dsadvantage of the global learnng les n that t may not make good use of the nformaton contaned n nstances near decson boundares. The dsadvantage of the local learnng les n that t may not make good use of the global nformaton contaned n the whole data set. The Nave Bayes classfer and the AODE algorthm belong to the global learnng, and the C4.5 algorthm and SVM belong to the local learnng. The Nave Bayes classfer and the AODE algorthm make good use of the global nformaton, but may not make good use of the local nformaton near decson boundares. The C4.5 algorthm and SVM make good use of the local nformaton, but may not make good use of the global nformaton. For the Nave Bayes classfer and the AODE algorthm, tranng the C4.5 algorthm and SVM n ther decson regons can make good use of the local nformaton contaned n nstances near decson boundares. For the C4.5 algorthm and SVM, beng traned n decson regons of the Nave Bayes classfer and the AODE algorthm can make good use of the global nformaton contaned n the whole data set. From sub-secton 5.2, t cannot mprove generalzaton abltes of the C4.5 algorthm and SVM by tranng the Nave Bayes classfer, the C4.5 algorthm and SVM n decson regons of the C4.5 algorthm and SVM, whch ndcates that t cannot obtan better generalzaton abltes to tran classfers n decson regons of the local learnng classfers by applyng the SD algorthm. The Tree algorthm can be regarded as tranng the Nave Bayes classfer n decson regons of the C4.5 algorthm and obtans better generalzaton ablty than both classfers (Kohav, 1996). However t dffers from usng the SD algorthm to tran the Nave Bayes classfer n decson regons of the C4.5 algorthm (SD(C4.5+)). Frst, the Tree algorthm trans the Nave Bayes classfer n regons generated by the C4.5 algorthm, not the real decson regons of C4.5 algorthm. Second, when tranng the C4.5 algorthm, the Tree algorthm takes test accuracy of the Nave Bayes classfer nto consderaton, whereas the SD algorthm does not consder ths factor. Thus the generalzaton ablty of the Tree algorthm s better than that of SD(C4.5+). The success of the Tree algorthm ndcates that t wll need much more complex methods to tran classfers n decson regons of classfers belongng to the local learnng.

11 Yan et al. / J Zhejang Unv-Sc C (Comput & Electron) n press 11 6 Conclusons Ths paper has nvestgated methods of tranng classfers n decson regons of the Nave Bayes classfer. The SD and three SD-soft algorthms all generate two-level classfer trees wth the Nave Bayes classfers as root nodes. Expermental results on 3 UCI data sets demonstrate the effectveness of these four new algorthms. When usng the C4.5 algorthm and SVM as leaf classfers, the SD algorthm can obtan better generalzaton ablty than the Nave Bayes classfer, leaf classfers, the Tree and the AODE algorthms. When appled twce and usng the Nave Bayes classfer as leaf classfers, the SD algorthm can obtan better generalzaton ablty than the Nave Bayes classfer and the Tree algorthm. When usng the C4.5 algorthm and SVM as leaf classfers, three SD-soft algorthms can obtan better generalzaton abltes than the SD algorthm, but argument selecton s crucal for these three SD-soft algorthms. The SD algorthm s also applcable for the AODE algorthm, but t s not applcable for the C4.5 algorthm or SVM. The SD and SD-soft algorthms can make good use of the nformaton contaned n nstances near decson boundares, and the nformaton may be gnored by global learnng classfers, such as the Nave Bayes classfer and the AODE algorthm. The SD and SD-soft algorthms can be regarded as a new method of generatng a hybrd of the global learnng and the local learnng. References Bezdek, J.C., Pattern Recognton wth fuzzy objectve functon algorthms. Plenum Press, New York. Bshop, C.M., 26. Pattern Recognton and Machne Learnng. Seres: Informaton Scence and Statstcs. Sprnger Verlag, New York, p Domngos, P., Pazzan, M., On the Optmalty of the Smple Bayesan Classfer under Zero-one Loss. Machne Learnng, 29: [do:1.123/a: ] Frank, E., Wtten, I.H., Generatng Accurate Rule Sets wthout Global Optmzaton. Ffteenth Internatonal Conference on Machne Learnng, Morgan Kaufmann, San Francsco, CA, USA, p Frosynots D, Stafylopats A, Lkas A., 23. A dvde-and-conquer method for mult-net classfers. Pattern Analyss and Applcatons, 6(1):32-4. [do:1. 17/s ] Huang, K.Z., Yang, H.Q., Kng, I., Lyu, M., 28. Machne Learnng: Modelng Data Locally and Globally. Sprnger-Verlag, New York Inc, p Kohav, R., Scalng Up the Accuracy of Nave-Bayes Classfers: A Decson-Tree Hybrd. Second Internatonal Conference on Knowledge Dscovery and Data Mnng, AAAI Press, p Mtchell, T.M., Machne Learnng. WCB/McGraw-Hll, p Murphy, P.M., Aha, D.W., UCI Repostory of machne learnng databases [ MLRepostory.html], Irvne, CA: Unversty of Calforna, Department of Informaton and Computer Scence. Pal, S.K., Mtra, S., Mult-layer perceptron, fuzzy sets and classfcaton. IEEE Transacton on Neural Networks, 3: [do:1.119/ ] Platt, J.C, Fast Tranng of Support Vector Machnes Usng Sequental Mnmal Optmzaton. In: Scholkopf, B., Burges, C., Smola, A. (Eds.), Advances n Kernel Methods: Support Vector Machnes. MIT Press, Cam-brdge, MA, p Qunlan, J.R., C4.5: Programs for Machne Learnng. Morgan Kaufmann, San Mateo. Qunlan, J.R., Improved use of contnuous attrbutes n C4.5. Journal of Artfcal Intellgence Research, 4: Vapnk, V.N., The Nature of Statstcal Learnng Theory. Sprnger, Berln Hedelberg. Vlasss, N. Lkas, A., 22. A Greedy EM Algorthm for Gaussan Mxture Learnng. Neural Processng Letters, 15(1): [do:1.123/a: ] Webb, G.I., Boughton, J., Wang, Z., 25. Not so nave Bayes: Aggregatng one-dependence estmators. Machne Learnng, 58:5 24. [do:1.17/s ] Wtten, I.H., Frank, E., 25. Data Mnng: Practcal Machne Learnng Tools and Technques, 2nd Edton. Morgan Kaufmann, San Francsco. Wu, X., Kumar, V., Qunlan, J.R., et al, 28. Top 1 Algo-rthms n Data Mnng. Knowledge and Informaton Sys-tems, 14(1):1-37. [do: 1.17/s ] Zheng, F., Webb, G.I., 25, A Comparatve Study of Sem-nave Bayes Methods n Classfcaton Learnng. Fourth Australasan Data Mnng Workshop, Unversty of Technology Sydney, p

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

The Research of Support Vector Machine in Agricultural Data Classification

The Research of Support Vector Machine in Agricultural Data Classification The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET 1 BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET TZU-CHENG CHUANG School of Electrcal and Computer Engneerng, Purdue Unversty, West Lafayette, Indana 47907 SAUL B. GELFAND School

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining the Optimal Bandwidth Based on Multi-criterion Fusion Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn

More information

Incremental Learning with Support Vector Machines and Fuzzy Set Theory

Incremental Learning with Support Vector Machines and Fuzzy Set Theory The 25th Workshop on Combnatoral Mathematcs and Computaton Theory Incremental Learnng wth Support Vector Machnes and Fuzzy Set Theory Yu-Mng Chuang 1 and Cha-Hwa Ln 2* 1 Department of Computer Scence and

More information

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan

More information

Fast Computation of Shortest Path for Visiting Segments in the Plane

Fast Computation of Shortest Path for Visiting Segments in the Plane Send Orders for Reprnts to reprnts@benthamscence.ae 4 The Open Cybernetcs & Systemcs Journal, 04, 8, 4-9 Open Access Fast Computaton of Shortest Path for Vstng Segments n the Plane Ljuan Wang,, Bo Jang

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Classifier Ensemble Design using Artificial Bee Colony based Feature Selection

Classifier Ensemble Design using Artificial Bee Colony based Feature Selection IJCSI Internatonal Journal of Computer Scence Issues, Vol. 9, Issue 3, No 2, May 2012 ISSN (Onlne): 1694-0814 www.ijcsi.org 522 Classfer Ensemble Desgn usng Artfcal Bee Colony based Feature Selecton Shunmugaprya

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for

More information

An Improvement to Naive Bayes for Text Classification

An Improvement to Naive Bayes for Text Classification Avalable onlne at www.scencedrect.com Proceda Engneerng 15 (2011) 2160 2164 Advancen Control Engneerngand Informaton Scence An Improvement to Nave Bayes for Text Classfcaton We Zhang a, Feng Gao a, a*

More information

Under-Sampling Approaches for Improving Prediction of the Minority Class in an Imbalanced Dataset

Under-Sampling Approaches for Improving Prediction of the Minority Class in an Imbalanced Dataset Under-Samplng Approaches for Improvng Predcton of the Mnorty Class n an Imbalanced Dataset Show-Jane Yen and Yue-Sh Lee Department of Computer Scence and Informaton Engneerng, Mng Chuan Unversty 5 The-Mng

More information

A User Selection Method in Advertising System

A User Selection Method in Advertising System Int. J. Communcatons, etwork and System Scences, 2010, 3, 54-58 do:10.4236/jcns.2010.31007 Publshed Onlne January 2010 (http://www.scrp.org/journal/jcns/). A User Selecton Method n Advertsng System Shy

More information

SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE

SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE Dorna Purcaru Faculty of Automaton, Computers and Electroncs Unersty of Craoa 13 Al. I. Cuza Street, Craoa RO-1100 ROMANIA E-mal: dpurcaru@electroncs.uc.ro

More information

Feature Selection as an Improving Step for Decision Tree Construction

Feature Selection as an Improving Step for Decision Tree Construction 2009 Internatonal Conference on Machne Learnng and Computng IPCSIT vol.3 (2011) (2011) IACSIT Press, Sngapore Feature Selecton as an Improvng Step for Decson Tree Constructon Mahd Esmael 1, Fazekas Gabor

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

An Evolvable Clustering Based Algorithm to Learn Distance Function for Supervised Environment

An Evolvable Clustering Based Algorithm to Learn Distance Function for Supervised Environment IJCSI Internatonal Journal of Computer Scence Issues, Vol. 7, Issue 5, September 2010 ISSN (Onlne): 1694-0814 www.ijcsi.org 374 An Evolvable Clusterng Based Algorthm to Learn Dstance Functon for Supervsed

More information

Implementation Naïve Bayes Algorithm for Student Classification Based on Graduation Status

Implementation Naïve Bayes Algorithm for Student Classification Based on Graduation Status Internatonal Journal of Appled Busness and Informaton Systems ISSN: 2597-8993 Vol 1, No 2, September 2017, pp. 6-12 6 Implementaton Naïve Bayes Algorthm for Student Classfcaton Based on Graduaton Status

More information

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION

BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION BAYESIAN MULTI-SOURCE DOMAIN ADAPTATION SHI-LIANG SUN, HONG-LEI SHI Department of Computer Scence and Technology, East Chna Normal Unversty 500 Dongchuan Road, Shangha 200241, P. R. Chna E-MAIL: slsun@cs.ecnu.edu.cn,

More information

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton

More information

The Shortest Path of Touring Lines given in the Plane

The Shortest Path of Touring Lines given in the Plane Send Orders for Reprnts to reprnts@benthamscence.ae 262 The Open Cybernetcs & Systemcs Journal, 2015, 9, 262-267 The Shortest Path of Tourng Lnes gven n the Plane Open Access Ljuan Wang 1,2, Dandan He

More information

Related-Mode Attacks on CTR Encryption Mode

Related-Mode Attacks on CTR Encryption Mode Internatonal Journal of Network Securty, Vol.4, No.3, PP.282 287, May 2007 282 Related-Mode Attacks on CTR Encrypton Mode Dayn Wang, Dongda Ln, and Wenlng Wu (Correspondng author: Dayn Wang) Key Laboratory

More information

Generalized Additive Bayesian Network Classifiers

Generalized Additive Bayesian Network Classifiers Generalzed Addtve Bayesan Network Classfers Janguo L and Changshu Zhang and Tao Wang and Ymn Zhang Intel Chna Research Center, Bejng, Chna Department of Automaton, Tsnghua Unversty, Chna {janguo.l, tao.wang,

More information

THE CONDENSED FUZZY K-NEAREST NEIGHBOR RULE BASED ON SAMPLE FUZZY ENTROPY

THE CONDENSED FUZZY K-NEAREST NEIGHBOR RULE BASED ON SAMPLE FUZZY ENTROPY Proceedngs of the 20 Internatonal Conference on Machne Learnng and Cybernetcs, Guln, 0-3 July, 20 THE CONDENSED FUZZY K-NEAREST NEIGHBOR RULE BASED ON SAMPLE FUZZY ENTROPY JUN-HAI ZHAI, NA LI, MENG-YAO

More information

Associative Based Classification Algorithm For Diabetes Disease Prediction

Associative Based Classification Algorithm For Diabetes Disease Prediction Internatonal Journal of Engneerng Trends and Technology (IJETT) Volume-41 Number-3 - November 016 Assocatve Based Classfcaton Algorthm For Dabetes Dsease Predcton 1 N. Gnana Deepka, Y.surekha, 3 G.Laltha

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System

Fuzzy Modeling of the Complexity vs. Accuracy Trade-off in a Sequential Two-Stage Multi-Classifier System Fuzzy Modelng of the Complexty vs. Accuracy Trade-off n a Sequental Two-Stage Mult-Classfer System MARK LAST 1 Department of Informaton Systems Engneerng Ben-Guron Unversty of the Negev Beer-Sheva 84105

More information

CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION

CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION 48 CHAPTER 3 SEQUENTIAL MINIMAL OPTIMIZATION TRAINED SUPPORT VECTOR CLASSIFIER FOR CANCER PREDICTION 3.1 INTRODUCTION The raw mcroarray data s bascally an mage wth dfferent colors ndcatng hybrdzaton (Xue

More information

Remote Sensing Textual Image Classification based on Ensemble Learning

Remote Sensing Textual Image Classification based on Ensemble Learning I.J. Image, Graphcs and Sgnal Processng, 2016, 12, 21-29 Publshed Onlne December 2016 n MECS (http://www.mecs-press.org/) DOI: 10.5815/jgsp.2016.12.03 Remote Sensng Textual Image Classfcaton based on Ensemble

More information

A Lazy Ensemble Learning Method to Classification

A Lazy Ensemble Learning Method to Classification IJCSI Internatonal Journal of Computer Scence Issues, Vol. 7, Issue 5, September 2010 ISSN (Onlne): 1694-0814 344 A Lazy Ensemble Learnng Method to Classfcaton Haleh Homayoun 1, Sattar Hashem 2 and Al

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

More information

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z.

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z. TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS Muradalyev AZ Azerbajan Scentfc-Research and Desgn-Prospectng Insttute of Energetc AZ1012, Ave HZardab-94 E-mal:aydn_murad@yahoocom Importance of

More information

Load Balancing for Hex-Cell Interconnection Network

Load Balancing for Hex-Cell Interconnection Network Int. J. Communcatons, Network and System Scences,,, - Publshed Onlne Aprl n ScRes. http://www.scrp.org/journal/jcns http://dx.do.org/./jcns.. Load Balancng for Hex-Cell Interconnecton Network Saher Manaseer,

More information

Machine Learning 9. week

Machine Learning 9. week Machne Learnng 9. week Mappng Concept Radal Bass Functons (RBF) RBF Networks 1 Mappng It s probably the best scenaro for the classfcaton of two dataset s to separate them lnearly. As you see n the below

More information

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1 A New Feature of Unformty of Image Texture Drectons Concdng wth the Human Eyes Percepton Xng-Jan He, De-Shuang Huang, Yue Zhang, Tat-Mng Lo 2, and Mchael R. Lyu 3 Intellgent Computng Lab, Insttute of Intellgent

More information

Classification / Regression Support Vector Machines

Classification / Regression Support Vector Machines Classfcaton / Regresson Support Vector Machnes Jeff Howbert Introducton to Machne Learnng Wnter 04 Topcs SVM classfers for lnearly separable classes SVM classfers for non-lnearly separable classes SVM

More information

Fast Feature Value Searching for Face Detection

Fast Feature Value Searching for Face Detection Vol., No. 2 Computer and Informaton Scence Fast Feature Value Searchng for Face Detecton Yunyang Yan Department of Computer Engneerng Huayn Insttute of Technology Hua an 22300, Chna E-mal: areyyyke@63.com

More information

A Clustering Algorithm for Chinese Adjectives and Nouns 1

A Clustering Algorithm for Chinese Adjectives and Nouns 1 Clusterng lgorthm for Chnese dectves and ouns Yang Wen, Chunfa Yuan, Changnng Huang 2 State Key aboratory of Intellgent Technology and System Deptartment of Computer Scence & Technology, Tsnghua Unversty,

More information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

Novel Fuzzy logic Based Edge Detection Technique

Novel Fuzzy logic Based Edge Detection Technique Novel Fuzzy logc Based Edge Detecton Technque Aborsade, D.O Department of Electroncs Engneerng, adoke Akntola Unversty of Tech., Ogbomoso. Oyo-state. doaborsade@yahoo.com Abstract Ths paper s based on

More information

Machine Learning. Topic 6: Clustering

Machine Learning. Topic 6: Clustering Machne Learnng Topc 6: lusterng lusterng Groupng data nto (hopefully useful) sets. Thngs on the left Thngs on the rght Applcatons of lusterng Hypothess Generaton lusters mght suggest natural groups. Hypothess

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

ISSN: International Journal of Engineering and Innovative Technology (IJEIT) Volume 1, Issue 4, April 2012

ISSN: International Journal of Engineering and Innovative Technology (IJEIT) Volume 1, Issue 4, April 2012 Performance Evoluton of Dfferent Codng Methods wth β - densty Decodng Usng Error Correctng Output Code Based on Multclass Classfcaton Devangn Dave, M. Samvatsar, P. K. Bhanoda Abstract A common way to

More information

A New Approach For the Ranking of Fuzzy Sets With Different Heights

A New Approach For the Ranking of Fuzzy Sets With Different Heights New pproach For the ankng of Fuzzy Sets Wth Dfferent Heghts Pushpnder Sngh School of Mathematcs Computer pplcatons Thapar Unversty, Patala-7 00 Inda pushpndersnl@gmalcom STCT ankng of fuzzy sets plays

More information

Unsupervised Learning

Unsupervised Learning Pattern Recognton Lecture 8 Outlne Introducton Unsupervsed Learnng Parametrc VS Non-Parametrc Approach Mxture of Denstes Maxmum-Lkelhood Estmates Clusterng Prof. Danel Yeung School of Computer Scence and

More information

Using Neural Networks and Support Vector Machines in Data Mining

Using Neural Networks and Support Vector Machines in Data Mining Usng eural etworks and Support Vector Machnes n Data Mnng RICHARD A. WASIOWSKI Computer Scence Department Calforna State Unversty Domnguez Hlls Carson, CA 90747 USA Abstract: - Multvarate data analyss

More information

Investigating the Performance of Naïve- Bayes Classifiers and K- Nearest Neighbor Classifiers

Investigating the Performance of Naïve- Bayes Classifiers and K- Nearest Neighbor Classifiers Journal of Convergence Informaton Technology Volume 5, Number 2, Aprl 2010 Investgatng the Performance of Naïve- Bayes Classfers and K- Nearest Neghbor Classfers Mohammed J. Islam *, Q. M. Jonathan Wu,

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned

More information

Complex System Reliability Evaluation using Support Vector Machine for Incomplete Data-set

Complex System Reliability Evaluation using Support Vector Machine for Incomplete Data-set Internatonal Journal of Performablty Engneerng, Vol. 7, No. 1, January 2010, pp.32-42. RAMS Consultants Prnted n Inda Complex System Relablty Evaluaton usng Support Vector Machne for Incomplete Data-set

More information

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms Course Introducton Course Topcs Exams, abs, Proects A quc loo at a few algorthms 1 Advanced Data Structures and Algorthms Descrpton: We are gong to dscuss algorthm complexty analyss, algorthm desgn technques

More information

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures

A Novel Adaptive Descriptor Algorithm for Ternary Pattern Textures A Novel Adaptve Descrptor Algorthm for Ternary Pattern Textures Fahuan Hu 1,2, Guopng Lu 1 *, Zengwen Dong 1 1.School of Mechancal & Electrcal Engneerng, Nanchang Unversty, Nanchang, 330031, Chna; 2. School

More information

Keywords - Wep page classification; bag of words model; topic model; hierarchical classification; Support Vector Machines

Keywords - Wep page classification; bag of words model; topic model; hierarchical classification; Support Vector Machines (IJCSIS) Internatonal Journal of Computer Scence and Informaton Securty, Herarchcal Web Page Classfcaton Based on a Topc Model and Neghborng Pages Integraton Wongkot Srura Phayung Meesad Choochart Haruechayasak

More information

Training of Kernel Fuzzy Classifiers by Dynamic Cluster Generation

Training of Kernel Fuzzy Classifiers by Dynamic Cluster Generation Tranng of Kernel Fuzzy Classfers by Dynamc Cluster Generaton Shgeo Abe Graduate School of Scence and Technology Kobe Unversty Nada, Kobe, Japan abe@eedept.kobe-u.ac.jp Abstract We dscuss kernel fuzzy classfers

More information

Automated Selection of Training Data and Base Models for Data Stream Mining Using Naïve Bayes Ensemble Classification

Automated Selection of Training Data and Base Models for Data Stream Mining Using Naïve Bayes Ensemble Classification Proceedngs of the World Congress on Engneerng 2017 Vol II, July 5-7, 2017, London, U.K. Automated Selecton of Tranng Data and Base Models for Data Stream Mnng Usng Naïve Bayes Ensemble Classfcaton Patrca

More information

X- Chart Using ANOM Approach

X- Chart Using ANOM Approach ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are

More information

Impact of a New Attribute Extraction Algorithm on Web Page Classification

Impact of a New Attribute Extraction Algorithm on Web Page Classification Impact of a New Attrbute Extracton Algorthm on Web Page Classfcaton Gösel Brc, Banu Dr, Yldz Techncal Unversty, Computer Engneerng Department Abstract Ths paper ntroduces a new algorthm for dmensonalty

More information

Three supervised learning methods on pen digits character recognition dataset

Three supervised learning methods on pen digits character recognition dataset Three supervsed learnng methods on pen dgts character recognton dataset Chrs Flezach Department of Computer Scence and Engneerng Unversty of Calforna, San Dego San Dego, CA 92093 cflezac@cs.ucsd.edu Satoru

More information

An Evaluation of Divide-and-Combine Strategies for Image Categorization by Multi-Class Support Vector Machines

An Evaluation of Divide-and-Combine Strategies for Image Categorization by Multi-Class Support Vector Machines An Evaluaton of Dvde-and-Combne Strateges for Image Categorzaton by Mult-Class Support Vector Machnes C. Demrkesen¹ and H. Cherf¹, ² 1: Insttue of Scence and Engneerng 2: Faculté des Scences Mrande Galatasaray

More information

SVM-based Learning for Multiple Model Estimation

SVM-based Learning for Multiple Model Estimation SVM-based Learnng for Multple Model Estmaton Vladmr Cherkassky and Yunqan Ma Department of Electrcal and Computer Engneerng Unversty of Mnnesota Mnneapols, MN 55455 {cherkass,myq}@ece.umn.edu Abstract:

More information

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1)

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1) Secton 1.2 Subsets and the Boolean operatons on sets If every element of the set A s an element of the set B, we say that A s a subset of B, or that A s contaned n B, or that B contans A, and we wrte A

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

Semi-supervised Classification Using Local and Global Regularization

Semi-supervised Classification Using Local and Global Regularization Proceedngs of the Twenty-Thrd AAAI Conference on Artfcal Intellgence (2008) Sem-supervsed Classfcaton Usng Local and Global Regularzaton Fe Wang 1, Tao L 2, Gang Wang 3, Changshu Zhang 1 1 Department of

More information

Audio Content Classification Method Research Based on Two-step Strategy

Audio Content Classification Method Research Based on Two-step Strategy (IJACSA) Internatonal Journal of Advanced Computer Scence and Applcatons, Audo Content Classfcaton Method Research Based on Two-step Strategy Sume Lang Department of Computer Scence and Technology Chongqng

More information

Improving Combination Methods of Neural Classifiers Using NCL

Improving Combination Methods of Neural Classifiers Using NCL Internatonal Journal of Computer Informaton Systems and Industral Management Applcatons. ISS 50-7988 Volume 4 (0) pp. 679-686 MIR Labs, www.mrlabs.net/jcsm/ndex.html Improvng Combnaton Methods of eural

More information

Learning from Multiple Related Data Streams with Asynchronous Flowing Speeds

Learning from Multiple Related Data Streams with Asynchronous Flowing Speeds Learnng from Multple Related Data Streams wth Asynchronous Flowng Speeds Zh Qao, Peng Zhang, Jng He, Jnghua Yan, L Guo Insttute of Computng Technology, Chnese Academy of Scences, Bejng, 100190, Chna. School

More information

A Semi-Supervised Approach Based on k-nearest Neighbor

A Semi-Supervised Approach Based on k-nearest Neighbor 768 JOURNAL OF SOFTWARE, VOL. 8, NO. 4, APRIL 03 A Sem-Supervsed Approach Based on k-nearest Neghbor Zhlang Lu School of Automaton Engneerng, Unversty of Electronc Scence and Technology of Chna, Chengdu,

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and

More information

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation Internatonal Conference on Logstcs Engneerng, Management and Computer Scence (LEMCS 5) Maxmum Varance Combned wth Adaptve Genetc Algorthm for Infrared Image Segmentaton Huxuan Fu College of Automaton Harbn

More information

Classifying Acoustic Transient Signals Using Artificial Intelligence

Classifying Acoustic Transient Signals Using Artificial Intelligence Classfyng Acoustc Transent Sgnals Usng Artfcal Intellgence Steve Sutton, Unversty of North Carolna At Wlmngton (suttons@charter.net) Greg Huff, Unversty of North Carolna At Wlmngton (jgh7476@uncwl.edu)

More information

Research of Neural Network Classifier Based on FCM and PSO for Breast Cancer Classification

Research of Neural Network Classifier Based on FCM and PSO for Breast Cancer Classification Research of Neural Network Classfer Based on FCM and PSO for Breast Cancer Classfcaton Le Zhang 1, Ln Wang 1, Xujewen Wang 2, Keke Lu 2, and Ajth Abraham 3 1 Shandong Provncal Key Laboratory of Network

More information

A Notable Swarm Approach to Evolve Neural Network for Classification in Data Mining

A Notable Swarm Approach to Evolve Neural Network for Classification in Data Mining A Notable Swarm Approach to Evolve Neural Network for Classfcaton n Data Mnng Satchdananda Dehur 1, Bjan Bhar Mshra 2 and Sung-Bae Cho 1 1 Soft Computng Laboratory, Department of Computer Scence, Yonse

More information

Parallel Implementation of Classification Algorithms Based on Cloud Computing Environment

Parallel Implementation of Classification Algorithms Based on Cloud Computing Environment TELKOMNIKA, Vol.10, No.5, September 2012, pp. 1087~1092 e-issn: 2087-278X accredted by DGHE (DIKTI), Decree No: 51/Dkt/Kep/2010 1087 Parallel Implementaton of Classfcaton Algorthms Based on Cloud Computng

More information

Network Intrusion Detection Based on PSO-SVM

Network Intrusion Detection Based on PSO-SVM TELKOMNIKA Indonesan Journal of Electrcal Engneerng Vol.1, No., February 014, pp. 150 ~ 1508 DOI: http://dx.do.org/10.11591/telkomnka.v1.386 150 Network Intruson Detecton Based on PSO-SVM Changsheng Xang*

More information

A Statistical Model Selection Strategy Applied to Neural Networks

A Statistical Model Selection Strategy Applied to Neural Networks A Statstcal Model Selecton Strategy Appled to Neural Networks Joaquín Pzarro Elsa Guerrero Pedro L. Galndo joaqun.pzarro@uca.es elsa.guerrero@uca.es pedro.galndo@uca.es Dpto Lenguajes y Sstemas Informátcos

More information

Tuning of Fuzzy Inference Systems Through Unconstrained Optimization Techniques

Tuning of Fuzzy Inference Systems Through Unconstrained Optimization Techniques Tunng of Fuzzy Inference Systems Through Unconstraned Optmzaton Technques ROGERIO ANDRADE FLAUZINO, IVAN NUNES DA SILVA Department of Electrcal Engneerng State Unversty of São Paulo UNESP CP 473, CEP 733-36,

More information

Problem Set 3 Solutions

Problem Set 3 Solutions Introducton to Algorthms October 4, 2002 Massachusetts Insttute of Technology 6046J/18410J Professors Erk Demane and Shaf Goldwasser Handout 14 Problem Set 3 Solutons (Exercses were not to be turned n,

More information

CS246: Mining Massive Datasets Jure Leskovec, Stanford University

CS246: Mining Massive Datasets Jure Leskovec, Stanford University CS46: Mnng Massve Datasets Jure Leskovec, Stanford Unversty http://cs46.stanford.edu /19/013 Jure Leskovec, Stanford CS46: Mnng Massve Datasets, http://cs46.stanford.edu Perceptron: y = sgn( x Ho to fnd

More information

An Image Fusion Approach Based on Segmentation Region

An Image Fusion Approach Based on Segmentation Region Rong Wang, L-Qun Gao, Shu Yang, Yu-Hua Cha, and Yan-Chun Lu An Image Fuson Approach Based On Segmentaton Regon An Image Fuson Approach Based on Segmentaton Regon Rong Wang, L-Qun Gao, Shu Yang 3, Yu-Hua

More information

Collaboratively Regularized Nearest Points for Set Based Recognition

Collaboratively Regularized Nearest Points for Set Based Recognition Academc Center for Computng and Meda Studes, Kyoto Unversty Collaboratvely Regularzed Nearest Ponts for Set Based Recognton Yang Wu, Mchhko Mnoh, Masayuk Mukunok Kyoto Unversty 9/1/013 BMVC 013 @ Brstol,

More information

CSCI 5417 Information Retrieval Systems Jim Martin!

CSCI 5417 Information Retrieval Systems Jim Martin! CSCI 5417 Informaton Retreval Systems Jm Martn! Lecture 11 9/29/2011 Today 9/29 Classfcaton Naïve Bayes classfcaton Ungram LM 1 Where we are... Bascs of ad hoc retreval Indexng Term weghtng/scorng Cosne

More information

From Comparing Clusterings to Combining Clusterings

From Comparing Clusterings to Combining Clusterings Proceedngs of the Twenty-Thrd AAAI Conference on Artfcal Intellgence (008 From Comparng Clusterngs to Combnng Clusterngs Zhwu Lu and Yuxn Peng and Janguo Xao Insttute of Computer Scence and Technology,

More information

A Deflected Grid-based Algorithm for Clustering Analysis

A Deflected Grid-based Algorithm for Clustering Analysis A Deflected Grd-based Algorthm for Clusterng Analyss NANCY P. LIN, CHUNG-I CHANG, HAO-EN CHUEH, HUNG-JEN CHEN, WEI-HUA HAO Department of Computer Scence and Informaton Engneerng Tamkang Unversty 5 Yng-chuan

More information

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images Internatonal Journal of Informaton and Electroncs Engneerng Vol. 5 No. 6 November 015 Usng Fuzzy Logc to Enhance the Large Sze Remote Sensng Images Trung Nguyen Tu Huy Ngo Hoang and Thoa Vu Van Abstract

More information

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements Module 3: Element Propertes Lecture : Lagrange and Serendpty Elements 5 In last lecture note, the nterpolaton functons are derved on the bass of assumed polynomal from Pascal s trangle for the fled varable.

More information

Journal of Chemical and Pharmaceutical Research, 2014, 6(6): Research Article. A selective ensemble classification method on microarray data

Journal of Chemical and Pharmaceutical Research, 2014, 6(6): Research Article. A selective ensemble classification method on microarray data Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(6):2860-2866 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A selectve ensemble classfcaton method on mcroarray

More information

Outline. Self-Organizing Maps (SOM) US Hebbian Learning, Cntd. The learning rule is Hebbian like:

Outline. Self-Organizing Maps (SOM) US Hebbian Learning, Cntd. The learning rule is Hebbian like: Self-Organzng Maps (SOM) Turgay İBRİKÇİ, PhD. Outlne Introducton Structures of SOM SOM Archtecture Neghborhoods SOM Algorthm Examples Summary 1 2 Unsupervsed Hebban Learnng US Hebban Learnng, Cntd 3 A

More information

Adjustment methods for differential measurement errors in multimode surveys

Adjustment methods for differential measurement errors in multimode surveys Adjustment methods for dfferental measurement errors n multmode surveys Salah Merad UK Offce for Natonal Statstcs ESSnet MM DCSS, Fnal Meetng Wesbaden, Germany, 4-5 September 2014 Outlne Introducton Stablsng

More information

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices Internatonal Mathematcal Forum, Vol 7, 2012, no 52, 2549-2554 An Applcaton of the Dulmage-Mendelsohn Decomposton to Sparse Null Space Bases of Full Row Rank Matrces Mostafa Khorramzadeh Department of Mathematcal

More information

Detection of hand grasping an object from complex background based on machine learning co-occurrence of local image feature

Detection of hand grasping an object from complex background based on machine learning co-occurrence of local image feature Detecton of hand graspng an object from complex background based on machne learnng co-occurrence of local mage feature Shnya Moroka, Yasuhro Hramoto, Nobutaka Shmada, Tadash Matsuo, Yoshak Shra Rtsumekan

More information

Case Mining from Large Databases

Case Mining from Large Databases Case Mnng from Large Databases Qang Yang and Hong Cheng Department of Computer Scence, Hong Kong Unversty of Scence and Technology, Clearwater Bay, Kowloon Hong Kong {qyang, csch}@cs.ust.hk http://www.cs.ust.hk/~qyang

More information

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero

More information

Concurrent Apriori Data Mining Algorithms

Concurrent Apriori Data Mining Algorithms Concurrent Apror Data Mnng Algorthms Vassl Halatchev Department of Electrcal Engneerng and Computer Scence York Unversty, Toronto October 8, 2015 Outlne Why t s mportant Introducton to Assocaton Rule Mnng

More information

INDEPENDENT COMPONENT ANALYSIS FOR NAÏVE BAYES CLASSIFICATION

INDEPENDENT COMPONENT ANALYSIS FOR NAÏVE BAYES CLASSIFICATION INDEPENDENT COMPONENT ANALYSIS FOR NAÏVE BAYES CLASSIFICATION FAN LIWEI (M.Sc., Dalan Unversty of Technology) A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY DEPARTMENT OF INDUSTRIAL & SYSTEMS

More information

A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION

A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION 1 THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMANIAN ACADEMY, Seres A, OF THE ROMANIAN ACADEMY Volume 4, Number 2/2003, pp.000-000 A PATTERN RECOGNITION APPROACH TO IMAGE SEGMENTATION Tudor BARBU Insttute

More information