Internatonal Journal of u- an e- Serce, Scence an Technology Trust Network an Trust Communty Clusterng base on Shortest Path Analyss for E-commerce Shaozhong Zhang 1, Jungan Chen 1, Haong Zhong 2, Zhaox Fang 1 an Jong Sh 1 1 Insttute of Electroncs an Informaton, Zheang Wanl Unersty, 315100, Nngbo,Chna 2 Insttute of Moern Logstcs, Zheang Wanl Unersty, 315100, Nngbo,Chna Dlut_z88@163.com, zh_1981@163.com Abstract Trust n e-commerce has become one of the most mportant ssues n onlne applcatons. Constantly, a user wll only search for the most creble of goos an serce proers an then take on ther transactons. How to confrm whch serce proers are the most truste for a user has become the most crtcal problems. Ths paper presents a trust network an trust communty clusterng for the analyss of the users most truste relatonshp. It uses the noes to represent the arous subects nole n the trust an use the connecton lnks to enote relatonshps. The weght of the lnks ncates the strength of the relatonshps. Frst, t construct a trust network agram whch has the weght alue of lnks, an then to analyze the clusterng propertes of the relatonshp accorng to the weghts an the path length. At last, t classfes the most truste subects to the same cluster for a user. Drect trust nformaton egree an global trust nformaton egree are use to ealuate trust relatons among subects an t ges an mproe shortest path algorthm to construct trust network. A clusterng algorthm base on coeffcent an path length s presente for E-commerce trust network communty. Experments show that the metho of bulng trust through the network moel can well escrbe the man nrect E-commerce trust an the algorthm has obous aantages n accuracy an tme cost. Keywors: E-commerce; trust nformaton egree; shortest path algorthm; trust communty clusterng 1. Introucton We can use the complex, nteracte network noe an network connectons to represent the nterests of the subects an the nteractons or relatonshp between them n E-commerce trust [1]. Many researchers hae stue the subect of the network moel of trust. Ths can help them to unerstan the subect of creblty an trust between subects. The results show that the trust relatonshp between subects to another can be regulate by trust network moel an support the functon of other's trust or creblty [2, 3, 4]. Trust network n e-commerce s a one of socal network. We can use the tool of socal network analyss to stuy the theory an methos. In the past, people hae been extensely stue on socal networks. Many scentsts hae stue the structure an propertes of the large an complex network [5, 6, 7]. Such as small-worl networks, collaboraton networks, mult-scale networks an communty networks n ther attrbutes analyss of the structure an the topology [8, 9, 10]. These networks an hae ery smlar characterstcs an propertes wth E-commerce trust networks. 31
Internatonal Journal of u- an e- Serce, Scence an Technology Most stues n the past hae focuse on the stuy of a sngle subect, or trust n the creblty of the establshment of the ssue. There s less n the network relatonshps of trust ssues between subects n E-commerce applcatons of real confence. For example, n the trust ealuaton moel of subect, the maorty of researches focus on the trust an creblty of the subect, an by the propertes an characterstcs of the nual to establsh trust an creblty for mpact on the surrounng assocaton to etermne the subect. Ths kn of trust an creblty s unchange n the network. For any subect n the network, ts trust egree s the same. Howeer, t s ffcult to trust a creblty whch establsh by unlateral confence ue to the uncertanty of the sources of nformaton an breath of the subect n e-commerce enronment. Trust relatonshp between subects s fferent n realty trust network. That s the trust egree s fferent that a subect to another subect compare to the subect to others. In aton, a rect trust egree between a subect an another oes not mean that the egree of global one between them. One coul ece whether to buy a prouct by some passe the ealuaton of other users though he has not nteract wth the merchant when he brows proucts n E-commerce. These users may hae he same characterstcs or propertes wth the user. The user may trust these users whch are n the trust network an then trust ther reew an purchase the goos. Therefore, a trust relatonshp of the trust an creblty n e-commerce enronment nees an ealuaton moel whch s establshe on self-trust relatonshp. It establshes a trust network moel of E-commerce by socal network analyss. A rect trust nformaton egree an a global trust nformaton egree are use to bul trust relatonshp among the subects n trust network. An mproe shortest path algorthm s use to bul trust network moel. It mplements trust communty clusterng analyss through the clusterng coeffcent an global trust nformaton egree an presents an mproe clusterng analyss algorthm for trust communty. 2. Socal Networks an Trust Network of E-commerce 2.1. Socal Network Analyss The research of relatonshp of socal network can be trace back to the 1960s n the fel of socology. Mlgram foun the characterstcs of small worl n socal network analyss [11]. Snce then, many researchers mae an extense research on socal network structure an ther characterstcs. Wth the eelopment of computer scence an network technology, t s growng n popularty n computer scence. In the relate research n the socal network, the graph moel s a ery mportant moelng tool. It has been abstracte an nual nto a noe an the relaton between the nuals nto the lnk. An then, a graph structure s bult. Through the stuy of ths partcular graph can analyss an mnng the nternal pattern an nformaton that be mple. Graph moel can be apple n socology, human behaor, transmsson of sease an nformaton an communcaton aspects of the Internet an other onlne communtes. Recently, some researches of ata mnng an structure mnng techncal hae proe that these networks hae the small worl propertes an characterstcs. In orer to mproe the effcency an scalablty of SA-Cluster, Zhou proposes an effcent algorthm Inc-Cluster to ncrementally upate the ranom walk stances gen the ege weght ncrements [12]. An Wu proposes a framework of an exact soluton an an approxmate soluton for computng rankng on a subgraph. He proe that the IealRank scores for pages n the subgraph conerge an analyze the stance between IealRank scores an ApproxRank scores of the subgraph [13]. 32
Internatonal Journal of u- an e- Serce, Scence an Technology 2.2. Trust Network of E-commerce In e-commerce enronment, Busness subects, nclung consumer an busness, ther trust an creblty has become an mportant ssue whch affects the eelopment of E-commerce. There are trust relatonshps exsts between consumers to consumers, busnesses to busnesses, an consumers to busnesses. These relatonshps rectly affect a user whether to trust qualty reew of another user of a certan goos, busness or serce. An t rectly mpact on the user s etermnaton for the qualty of busness, goos or serces. An further affect the user s choce of goos an serces. We can ealuate the trust relatonshp of subect through socal network analyss. Golbeck propose a trust nference mechansm for trust relaton establshment between a source partcpant an the target one base on aeragng trust alues along the socal trust paths [14]. G. Lu put forwar a framework of trust propagaton to stuy the complex socal network by the path selecton problem an a new concept Qualty of Trust s use to guarantee a certan leel of trust worthness n trust propagaton along a socal trust path [15]. R. James examne the role of trust from arous aspects wthn telemecne, wth partcular emphass on the role that trust plays n the aopton an aaptaton of a telemecne system [16]. 3. Constructon of Trust Networks 3.1. Trust Informaton Degree In a trust network, the leel of trust between one noe to another can be expresse by metrc functon of trust. Trust egree s the theoretcal bass for ealuaton moel of a trust network. A well-esgne trust egree plays a tal role n the accuracy of assessment for trust relatonshps n network. In ths paper, we ntrouce mutual nformaton functon as a measure of confence n the assessment of the trust network moel. Mutual nformaton s use to ncate certan nformaton that sent or recee some nformaton an brought out some arable nformaton n nformaton theory. If there are two noes whch hae a trust relatonshp n the trust network, you can get some nformaton about another noe when the nformaton of the one noe s recee. Ths nformaton can be measure by mutual nformaton. Therefore, t can etermne trust relatonshp through mutual nformaton between two noes. An t can measure the sze of the trust. Defnton 1trust nformaton egree): For a rust network space, there are G V, S).where V s the set of noes n all the space. S s the set of lnks n all the space. Let an are two fferent noes on the space. The trust nformaton egree s efne as Bel, ). Bel, ) = p, ) p, ) log p ) p ) 1) From the trust nformaton egree, we can fn the stance between the ont probablty ensty functon of an on oman U, that s p, ),an p ) p ), reflects the sze of mutual nformaton of the arables. When the noe 33
Internatonal Journal of u- an e- Serce, Scence an Technology an nepenent each other, there s p, ) = p ) p ).ths represents that the stance between p, ) an p ) p ) s zero. An then the mutual trust nformaton egree s zero;when there s trust relatonshp between an,that s when p, ) p ) p ),an then the stance between p, ) an p ) p ) s not zero. Furthermore, the more the stance, the less the mutual nformaton. An the trust nformaton egree s larger between an.in ths case, the trust relatonshp of an s more obous. 3.2. Drect Trust Informaton Degree Defnton 2rect trust nformaton egree): In a trust network, f there are hstory relatonshps between two noes an, an the two noes contonal nepenent each other, then the trust nformaton egree can be obtane by equaton 1). Ths kn of trust egree s call recte trust nformaton egree, label as Bel, ). Drect trust nformaton egree s assesse by computng rectly nteract recore hstory of the noes an the others. We use the total number of hstorcal nteracton, the number of successful nteracton, the number of falures an the probablty relatonshp between ther mutual nteractons as the bass for rect creblty. Assume that the noe an. The falures for noe an A, s the total number of the rect nteracton at a tme nteral for T, represents the number of successes an F,.we use A,, T, an F, the number of to construct rect trust nformaton egree for noe to.the egree s represents by Bel, ) an the Bel, ) s the rect trust nformaton egree functon of noe to. There s a rect trust nformaton egree between any two noes n a trust network. The rect trust nformaton egree between them s zero when the rect nteracton hstory s empty between the two noes. The calculaton algorthm s as follows. Algorthm 1: Calculaton the rect trust nformaton egree of any two noes Step 1: Select any one noe,count the nteracton hstory of t to another noe an calculate functon A,, T, an F,. Step 2: If A 0, then calculate the trust nformaton egree through equaton, 1).that s:, ) p, Bel = p, ) log p ) If A 0, then let Bel, ) 0., Step 3: Repeat wth the next noe an untl hae searche all network noes through space. ) p ) 34
Internatonal Journal of u- an e- Serce, Scence an Technology 3.3. Globe Trust Informaton Degree Drect trust nformaton egree can be calculate by rect nteracton hstory between two noes. But f the hstory of ther nteracton s none, then the rect trust alue s zero. Howeer, ths oes not mean that these two noes trust relatonshp oes not exst. Ths trust alue can be passe through other noes. Ths kn of trust whch s obtane by pass through others s global trust nformaton egree. Global trust nformaton egree s to conser the oerall egree state of trust. Trust between two noes s netable affecte by the rect an nrect nformaton that of other noes n network. Thus ths effect wll affect the relatonshp of trust between the two noes. In aton, f there s no rect connecton lnks between two noes, the trust nformaton egree between them wll be calculate out through trust alue of the ntermeate noes n the network. Defnton 3globe trust nformaton egree): the trust relatonshp between noe an can be assesse through global relatonshp n the whole network. Ths trust egree s efne as globe nformaton egree of noe an as Bel, ). a an be represente In ths paper, an mproe shortest path algorthm s aopte to calculate the globe trust nformaton egree. Shortest path algorthm s a typcal graph search algorthm that wely use n graph moel. The algorthm assesses the optmal path between noes by etectng eges weght connecte noes. The theory an methos of graph moel can be use to search the optmal path n fnng trust relatonshp network structure. The best path n here s no longer the shortest path, but s the optmal path for trust. It nclues multple noes n ths trust path. These noes can be look as the best relatonshp of trust from the source noe to estnaton noe. By analyzng the global trust nformaton egree of these noes to estnaton noe, the trust relatonshp that these noes to estnaton noe can be etermne. These trust nformaton status are the most mportant reference for source noes. The source user may most trust the cret ealuaton that those user are on the truste path who apprase some prouct or busness n E-commerce. Thus t wll affect the source user to purchase target goos or serces. Apparently, the source user has no rect nteracton wth target goos or busness. Ths mechansm soles the problems of hang no way to trust an to etermne for user when he wants to purchase some goos facng the arety of user ealuaton. The metho proes a rght feasble soluton for the user to choose a aluable, truste an obecte ealuaton. The exstng search algorthm of research networks for shortest path most focuse on the analyss lnks of bulng structure. An less for the relatonshp between the noes of the search problem. Especally at the tme of the absence of a rect connecton between two or more enttes to bul a network. In aton, the foun path shoul be reflectng the fact that the most closely lnke between noes n orer to proe the most mportant fact. The typcal relatonshp searchng algorthm s breath-frst search methos n the exstng analyss. Howeer, the algorthm can not fn the closest relatonshp between noes. It s often lmte to rect search an ffcult to achee the global search there s no rect contact or nrect relatonshp. To ths en, we mproe the basc BFS algorthm n the paper. 35
Internatonal Journal of u- an e- Serce, Scence an Technology Algorthm 2: An mproe BFS algorthm Input: The source noe s an termnal trust noe t,the rect trust nformaton egree Bel, ) of Between any two noes n the space. Output: The globe trust nformaton egree of source trust noe s to termnal trust noe t, Bel a s, t). Step 1: Let T s a set of a tree an T {s} ; Step 2: Select a noe an let by algorthm 1. s,calculate the rect trust egree Bel s, ) Step 3: If the alue Bel s, ) Less than a threshol alue, that s Bel s, ),then Dscar the noe an go to Step 2.If there s Bel s, ),then to the next step. Step 4: If the noe s n set T then go to Step 2. If the noe s not n set T Then a the noe n the tree T an set the noe s as the parent noe of,that s let T T an Pa s. Step 5: for each noe whch s on the path from s tot : ⑴.If the alue of nformaton egree s s the Maxmum, then set the path nto s. ⑵. If the noe s an Intermeate noe on the path an t s on the path of s t,then elete the noe an lnk ts parent noe an subnoe. ⑶. If the path t s on the path of s t, then set the path nto t untl there s no noe on the path s whch s on the path t. If the resultng tree noes whch hae a path set are a subset of another set of noes of a path, then elete the path. An then set the leaf noe nto one noe. ⑷.calculate the trust nformaton egree alue of the path. Bel a 1 2 t s, t) Bel s, ) Bel s, ),..., Bel, ) 2) 4. Clusterng Analyss of Trust Communty The purpose of the trust communty clusterng n the trust network s e nto a hgh egree of mutual trust communty from the trust subects. In these communtes, the man boy of each trust subects has a hgh egree of global trust nformaton. They can share ealuaton of the goos or serces among the subect n a same communty an share the experence egree of E-commerce. Ths help people to remoe nterferng factors n e-commerce enronment an extract the most aluable nformaton. 36
Internatonal Journal of u- an e- Serce, Scence an Technology 4.1. Clusterng Coeffcent Watts an Strogatz use clusterng coeffcent to escrbe the network noe connecton egree n small-worl network analyss. In fact that the clusterng coeffcent can also be use n terms of small-worl network or mult-scale network. We can use t to escrbe the characterstcs of network structure n other complex network analyss. Clusterng coeffcent represents the closeness of a noe wth other noes on behalf of the network. It enotes the egree of trust n E-commerce network between the busness subects. For noe wth a k egree k enotes there are k connecte eges), ts clusterng coeffcent can be efne as: Clusterng coeffcent C : t s a measurement parameters of the closely egree of neghbor noes. C enotes the rato of actual number of eges of subgraph to that wth the largest number of eges: C k 2t k 1) k s the number of neghbor noes. Let C enotes the mathematcal expectaton of C of all noes, an then the expectaton s the clusterng coeffcent: C C n 1 n Clusterng coeffcent escrbes tghtness hol together of the noes on the network. It s the local features of a network. Among them, n enote the number of ege that the noe connecte neghbor noes. 3) 4) 4.2. Clusterng Algorthm for Trust Communty In orer to establsh a hgh clusterng an hgh nformaton egree of trust network clusterng, we use the global trust nformaton egree as the ealuaton factor, whch s stance between two noes n the network. One fference wth the stance functon n the trust network s that the greater the egree of nformaton, the smaller the stance that the path s. Conersely, the smaller the egree of nformaton, the greater the stance. Clusterng of trust network s to achee some noes an lnks, whch mutual trust nformaton egree hae reache a threshol, nto a same communty. The clusterng wll be has a hgh clusterng, hgh-trust propertes. Defnton 4: For a trust network, a sub-graph s a clusterng of subects of E- commerce, whch s obtane by eletng m arcs from the regular network so that maxmzng f abel a bc, when a regular network G V, S) wth k egree s m gen. In whch V s subects noes set an S s arcs or lnks set. In contonal parameter f abel a bc m, a an b are constants, m s nteger, Bel a an C s characterstc path length an clusterng coeffcent respectely. Solng the optmal connectty problems of the noes n ths network s a NP problem. We propose an algorthm for the optmzaton as follows. 37
Internatonal Journal of u- an e- Serce, Scence an Technology Specfc clusterng algorthm for truste communty network s as follows: Algorthm 3: Trust Communty clusterng Step 1. Repeat cut an arc, whch coul maxmze f, untl m arcs are moe. Step 2. Jon an arc that coul maxmze f an make a ugment. If the one arc s the same as the cut one then the algorthm wll en. Step 3.cutng an arc, whch can maxmze f an then ump to Step 2. The parameters Bel a an C that satsfy the maxmal f max are the clusterng about the noes, n whch the clusterng group can be expresse as follow: ' V { V Bel } 5) 5. Experment an Analyss a P. Massa [17] use a large onlne communty Epnons ata sets to ealuate the trust relatonshp. We also use the ata set to analyss the performance of clusterng trust communty. Epnons ata set s a who-trust-whom onlne socal network of a general consumer reew ste Epnons.com. Members of the ste can ece whether to trust each other. All the trust relatonshps nteract an form the Web of Trust whch s then combne wth reew ratngs to etermne whch reews are shown to the user [18]. The ata set conssts of two parts, the ratng_ata sets an trust_ata sets. Ratng_ata sets nclue three tems, the user_, tem_, ratng_alue, an comprsng about 49,290 user s noes, ratng of 139,738 tems. Trust_ata sets consst of source_user_, target_user_, trust_statement_alue, an comprsng 49,290 user noes trust status. We let the ratng ata set as tranng ata an use algorthms 1 to 3 to bul trust communtes. The trust_ata sets trust looke as a test set to assess the accuracy of the results. In orer to test our algorthms we use other two fferent algorthms to compare them. The frst algorthm s a stanar Collaboratng Flterng one an the secon s Mole Trust [17]. The alty of communty clusterng of trust network can be ealuate by the accuracy an tme effcency. When the trust communty cluster that noe nclue s n the consstency wth hgher confence of trust_ata sets, t ncatng that the noes n the communty cluster hae close relatonshp of trust wth other noes n the same communty an t shows the trust communty clusterng s correct. When there s low confence n the consstency of trust_ata sets, t shows the trust communty clust erng s ncorrect. In ths case, noes n the same trust communty cluster exst not a hgh trust relatonshp. The Accuracy s efne as: Nc Accuracy N N c Where Nc s the number of correct noes an N the ncorrect. The results show n graph 1 to 3. The horzontal lnes represent the number of noes n the test samples set; 6) 38
Internatonal Journal of u- an e- Serce, Scence an Technology the ertcal lne represents the accuracy of clusterng communty. Comparson of three methos shown n Fgure 1. Fgure 1. The Tren Lne of Accuracy for the Three Algorthms In the trust before the 10000 test samples, the stanar Collaboratng Flterng relately has a low accuracy of the algorthm. For the other two algorthms they hae a hgher accuracy of the results, n whch the Mole Trust reache 0.6 at 9000 samples an communty clusterng can achee 0.7. The Mean Absolute Error s also use to analyss accuracy also. The horzontal lne ncates the number of sample noes; ertcal lne represents the MAE, a comparson of the three methos shown n Fgure 2. Fgure 2. The MAE of the Three Algorthms uner Dfferent Number of Tranng Noes From aboe we can see that there s a certan nfluence on the results from the number of tranng samples. When the tranng samples are low, the errors are hgher. When the number of samples for more than 45,000, the error s sgnfcantly reuce. By selectng the fferent threshols to test the conergence of the algorthm. Take 0.6, 0.8 an 0.10, respectely, whch to analyze the algorthm's executon tme. The results shown n Fgure 3. 39
Internatonal Journal of u- an e- Serce, Scence an Technology Fgure. 3. The Secons of Conergence uner Dfferent Threshol 6. Concluson A trust network moel n E-commerce by socal network analyss s establshe n ths paper. Ths moel combnes rect trust nformaton egree an global trust nformaton egree. It has an aantage of bulng a trust relatonshp network between the subects. It proposes an mproe shortest path algorthm to bul trust network moel. It propose the concept of trust communty networks an through communty clusterng analyss to construct trust relatonshp. It also ges the algorthms for the global trust nformaton egree an trust communty clusterng n E-commerce. The experments show that the metho of bulng trust network moel can well escrbe the man nrect trust n E-commerce an the algorthms has obous aantages n accuracy an n tme cost. Acknowlegements Ths work was supporte by the Natonal Natural Scence Founaton of Chna Grant No. 71071145 an 61003254), the Natural Scence Founaton of Zheang Pronce Grant No.Y6090027, Y1101123, an Y1110200 ) an the Moern Port Serce Inustry an Culture Research Center of the Key Research Base of Phlosophy an Socal Scences of Zheang Pronce. References [1] Joseph AC, Benamn BMS an Robert DSL, Sharng Informaton an Bulng Trust through Value Congruence, Inf Syst Front, ol. 9, 2007), pp. 515 529. [2] Arnam N, Trust Dren Informaton Sharng n Peer-to-Peer Socal Networks: Desgn an Analyss, Doctor of Phlosophy Unersty of Mantoba, 2008). [3] Abul R an Hales S, Usng Recommatons for Managang Trust n Dstrbute Systems, In IEEE Malaysa Internatonal Conference on Conmmuncaton, 1997) Noember. [4] John OD an Barry S, Trust n Recommener Systems, IUI 05, San Dego, Calforna, USA, 2005) January, pp. 167-174. [5] Gran M an Newman MEJ, Communty structure n socal an bologcal networks, Proc. Natonal Acaemy of Scences of the Unte States of Amerca, ol. 99, no. 12, 2002), pp. 7821-7826. [6] Newman MEJ, Clusterng an preferental attachment n growng networks, Phys. Re. E, ol. 64, 2001). [7] Barabas AL an Albert R, Emergence of Scalng n Ranom Networks, Scence, ol. 286, no. 5439, 1999), pp. 509-512. 40
Internatonal Journal of u- an e- Serce, Scence an Technology [8] Barabas AL, Jeong H, Raasz R, Na Z, Vcsek T an Schubert A, On the topology of the scentc collaboraton networks, Physca A, ol. 311, 2002), pp. 590-614. [9] Racch P, Castellano C, Ceccon F, Loreto V an Pars D, Denng an entfyng communtes n networks, Proc. Natl. Aca. Sc. USA, ol. 101, 2004), pp. 26-58. [10] Newman MEJ, Bara AL an Watts DJ, The Structure an Dynamcs of Networks, Prnceton Unersty Press, 2006). [11] Mlgram S, The small worl problem, Psychology toay, ol. 2, 1967), pp. 60-67. [12] Zhou Y, Cheng H an Yu JX, Clusterng Large Attrbute Graphs: An Effcent Incremental Approach, 2010 IEEE Internatonal Conference on Data Mnng, 2010), pp. 689-698. [13] Wu Y an Rasch L, Approxrank: Estmatng rank for asubgraph, n ICDE, 2009), pp. 54 65. [14] Golbeck J an Henler J, Inferrng trust relatonshps n web base socal networks, ACM Transactons on Internet Technology, ol. 6, no. 4, 2006), pp. 497 529. [15] Lu G, Wang Y an Orgun M, Qualty of trust for socal trust path selecton n complex socal networks, In AAMAS 10, 2010). [16] James R, Trust an Trustworthness: A Framework for Successful Desgn of Telemecne, Doctor of Phlosophy. Noa Southeastern Unersty, 2010) June. [17] Massa P, Trust-aware Recommener Systems, RecSys 07, Mnneapols, Mnnesota, USA, 2007) October 19 20. [18] http://www.trustlet.org/wk/downloae_epnons_ataset,2011.12 Authors Shaozhong Zhang, Ph.D., professor. Research fellow of Electronc Serces Research Center of Zheang Unersty an Nngbo Insttute of Electronc Serces. Expert commttee of Chna Insttute of Communcatons of clou computng an SaaS. Execute rector of the Electronc Commerce Assocaton of Nngbo. Apprasal experts of the Natonal Natural Scence Founaton of Chna. As a proect leaer, prese oer a number of research proects of the country, the Mnstry of Eucaton, Zheang Pronce an Nngbo Cty. Jungan Chen has recee hs master egree n computer scence from the Zheang Unersty of Technology, Chna n 2005. He s an assocate professor n Zheang Wanl Unersty now. Artfcal Immune System apple to computer securty s hs man research recton. Haong Zhong has recee hs octoral egree n cartography an geographc nformaton system from the East Chna Normal Unersty n 2011. Durng the octor stuy tme he has partcpate n one Chna Natonal Natural Scence Founaton proect Grant No. 41001270) an 5 proncal applcaton proects manly responsble for the mplementaton of these proects). In July 2011, he one n the Moern Logstcs School of Zheang Wuanl Unersty as a lecturer. Now hs maor research nterests nclue: e-commerce trust, moble e-commerce applcaton, LBS an so on. 41
Internatonal Journal of u- an e- Serce, Scence an Technology Zhaox Fang recee the BEng n communcaton engneerng an the PhD n electrcal engneerng from Fuan Unersty, Shangha, Chna, n 2004 an 2009, respectely. Snce June 2009, he has been an assstant professor n the School of Electronc an Informaton Engneerng, Zheang Wanl Unersty, Nngbo, Chna. Hs research nterests nclue terate etecton, frequency oman equalzaton, an cooperate communcatons. Jong Sh recee the BEng n communcaton engneerng from Zheang Unersty of Technology n 2005, an the PhD n electrcal engneerng from Beng Unersty of Posts an Telecommuncatons n 2010. Snce Sep. 2010, he has been a lecturer n the School of Electronc an Informaton Engneerng, Zheang Wanl Unersty, Nngbo, Chna. Hs research nterests nclue nformaton theory, OFDM systems, an short-range wreless communcaton technologes. 42