Research On Peer Producer Selection Behavior And Its Re-weighted Guiding Mechanism

Using Matthew Effect, Kleinberg proposed two re-weighted mechanisms for two specialcases of scientific community project game. The new mechanisms could maximize thecommunity benefit. In this paper, we keep our focus on the advance of Kleinberg re-weightedmechanism in peer production community. Then we show the guiding model for the re-weightmechanism of the selection of producers.For the general model of the scientific community project game, only re-weight theparticipants or the projects may not make the optimal assignments the Nash equilibrium. Inthis paper, we propose the generalized assignment&project game model which is theadvance of the Kleinberg re-weighted mechanism. Its objective function is maximizingcommunity benefit and its constraint is that the generalized assignment is the Nashequilibrium of participants non-cooperation game. It is an NP problem to solve generalizedassignment&project game model. In this paper, we try to find its satisfactory solution andpresent the greedy algorithm and advanced greedy algorithm.Through simulation, we found that the greedy algorithm could find the good satisfactorysolution of the generalized assignment&project game model, which is better than thatobtained by simulated annealing algorithm. The solution obtained by advanced greedyalgorithm based on regret value is better than that obtained by greedy algorithm.Peer Production is a new kind of production mode using online collaboration andproducing knowledge product by Internet users all over the world. These Internet users arecalled peer producers. Open source community and Baike on line are typical community ofpeer production. In order to study the Kleinberg re-weighted advance mechanism and guidingmodel for producer selection behavior, the selection behavior of peer producers should befirstly studied. In this paper, we study the selection behavior of entry producer, taking BaiduBaike as example.Firstly, we classify the Baike entries according to domain, type and editing style and create the map from entries set to domain set, type set and editing style set, respectively. Thenwe carry out the following two works on the producer entry selection serial:(1) analyze thevariation of producer selection frequency based on entry domain, entry type and entry editingstyle;(2) using the L-Z complexity and information entropy, we analyze the complexity of thethree serials. We found that entry producer has personal preference for every entryclassification, and complexities of producer selection behavior from information entropy andL-Z complexity are closely fit.Secondly, using the three types of entries we further define time sequence of interval andtime for the same entry field, type and edit style. Then we conduct the following study.(1)L-Z complexity and information entropy are used to study complexity of the sequences.(2)We analyze the power-law distribution characters of the two types of time serials, and findthat the two types of time serials of the three classification follow power-law distributionwhen the length of the sequence more than1000.(3) We use the way of MF-DFA to analyzethe multi-fractal property and sequence of interval and time. And we find that most of timesequences have a long-term memory, and have the multi-fractal property for those with thelength of the sequence more than1000.Finally, according to peer production community selection behavior character, wepropose two types Kleinberg re-weighted advance mechanisms and their guiding models, rthpower re-weight mechanism based on expected benefit and its guiding model, and re-weightmechanism based on introduce index and its guiding model. Using Baidu Baike entryproducer selection behavior character on entries type, we carry out simulation on the guidingmodel. The simulation proves that the model based on introduce model could easily describethe guiding mechanism of selection behaviors of entry producers.